Tweet
I finally got around to reading the thread on linear workflow. It contained a lot of good information but it also may have caused a lot of confusion for some. I believe this is because it lacked a clear definition of the target audience. Had Throb defined the type of work he does and his workflow up front, his tutorial and thread would have been a lot clearer.
The intent of this article is to point out that a linear workflow is not always necessary nor is it always the best solution. I know that there are many people whose work requires a different approach than that which I outline below. But since I’ve had great success with this process, I thought I’d share it in hopes that there are some who will benefit by it.
My work consists mostly of creating architectural presentations shown to a non-architect audience. My mandate is to produce the most photorealistic images possible. The initial presentations are projected electronically whereas the final presentation consists of 32 x 40 inch prints. I use 3D Studio Viz, Vray 1.47.03 and PhotoShop CS2 on WinXP pro.
Because of my workflow requirements, I have decided that using gamma 2.2 is better for me than 1.0 or linear. There are several reasons for this. First, most of the many hundreds of monitors in our organization are not calibrated. Those that are have been calibrated to gamma 2.2. Of course, I could always apply the appropriate profile before distributing my images, but I don’t like having to keep track of different versions of my files. Plus the process adds another step that takes time.
Second, most monitors perform much better at 2.2 than at 1.0 and gamma 2.2 images are displayed more accurately, even though the file itself might be limited due to the gamma curve. Also, calibrating a Windows PC system to 1.0 makes the desk top so washed out it is really uncomfortable to use, even when using dark desktop colors.
Third, most of our peripheral devices such as projectors, printers, scanners and plotters are intended to output 2.2 images. In tests I’ve performed, I’ve found that there is no visual difference in output quality between linear images vs. gamma 2.2 images. Even the PhotoShop tests described on the AIM website http://www.aim-dtp.net./aim/evaluation/gie/index.htm don’t seem to have as big an impact as claimed. In short, the added steps needed to deal with 1.0 images in my workflow just aren’t justified.
While I understand the logic behind a linear workflow, I’ve found that using a consistent 2.2 process together with good color profile management and proper monitor calibration makes up for any reduction in file integrity caused by the 2.2 gamma.
Workflow
My work typically begins with architectural massing and volume models that I render using monochromatic colors, usually just tonal variations of gray and white. I use values around the 130 – 140 range. Accurate lighting is important at this stage as it helps define the architectural forms and gives depth. For interiors I use slightly cool plane lights at window openings and slightly warm sphere lights at fixtures, both with a multiplier of 1 ~ 2. For exterior work, I use a target direct light with a slightly warm color, multiplier 1, and a slightly cool environment skylight, multiplier .35. The second stage involves importing scanned images as texture maps, tweaking the lighting, adding furnishings and fixtures for interior images and landscape and entourage for exterior images. The third step is to make final lighting and material tweaks and create the final renderings. The final step is to bring the image into PhotoShop, tweak colors once more, add a few blending layers for softening effects and print.
Here’s how I set up my system and software:
Monitor calibration
Many hardware devices are available and people in our organization use them frequently. However, I’ve gotten virtually the same results using Adobe Gamma and the calibration image and procedures found here:
http://www.aim-dtp.net./aim/calibration/index.htm
and here:
http://www.aim-dtp.net./aim/calibrat...d_contrast.htm
Scanner calibration
I use a flatbed scanner for most of my texture capturing but occasionally use a digital camera. In either case the input device must be properly set up and an accurate profile assigned. The profile is created by comparing a scan or digital photograph of a Color Target to the LAB colors that were used to create the target. Any discrepancy is recorded and a formula or Profile created to correct for those differences. Information can be found here:
http://www.coloraid.de/
In my scanner setup, I select output to gamma 2.2. Comparisons between 2.2, 1.0 and non corrected scans favor 2.2. Of course, this might be different for other scanners.
PhotoShop Color Management
The key to the whole process is using appropriate color settings and profiles. In the color settings dialog you will see a whole bunch of choices. I know there is debate on which settings to use, but my preference is to use Adobe RGB (199
for the working RGB space and US Web Coated (SWOP) v2 for the CMYK space. I prefer Adobe RGB because it has a gamut that is more encompassing than the smaller gamuts like AppleRGB, ColorMatch RGB or sRGB and is more closely aligned to the NTSC gamut. As for the CMYK color space, the SWOP v2 space gives me slightly better print results on my HP 5500 plotter than using the Adobe RGB setting. Since I allow my Printer to Determine Colors (selected in the print dialog box) I make sure that Color Management is turned on in the printer driver and that the printer firmware is selected to the appropriate CMYK profile.
The other Color Settings in PhotoShop are as follows: Mismatched and Missing Profiles: set all to Ask When Opening. Engine: Adobe (ACE). Intent: Relative Colorimetric. Use Black Point and Dither: both checked. Desaturate and Blend: unchecked.
3D Studio Settings
In Preference Settings I check Enable Gamma Correction and set my Display and Bitmap Gamma all to 2.2. If I get a bitmap file that is different than 2.2, I override it in the load bitmap dialog box.
For color mapping, I use Linear Multiply, Dark1.0, Light1.0 & Clamp Output unchecked. I’ve tested different settings and these work best for my workflow. Using Gamma Correction Dark1.0, Light .455 and saving the file with gamma 1.0 comes close but textures come out a bit too dark, colors over-saturated and for some reason the AA is not as good. (To compensate for the 1.0 output, I assigned the AIM RGB Trinitron D65 G1.0 profile. Assigning other 1.0 profiles had a similar effect).
Practical application of the settings
Since most of my textures and materials exist in the real world, I can compare the scanned image and the rendered 3D image with the actual physical object. This provides a good test to the accuracy of the calibrations and settings.
When I open a scanned file or digital image, PhotoShop asks what I want to do. Since I created a scanner profile, I select Assign Profile then select the profile that I created and check Convert Document to Working RGB. Since my monitor is calibrated accurately and I’ve created a good scanner profile, what I see on my screen should look very close to the actual object. When I save it, I make sure that the Adobe RGB profile is saved with it by checking the ICC Profile check box in the Color section of the Save As dialog box.
When I import the texture map into 3D Studio, I make sure it looks the same in the Material Editor as it does in real life, another good check for accuracy. When I render the scene, I try to set up my lights so that the rendered material looks close to the real object in tone and color. Only after I’m satisfied with the exposure and color of the texture do I adjust the lighting for mood and effect. When I save the rendered file and load it into PhotoShop, I again compare the result with the actual materials that I scanned to make sure I’m not faking things too much with lighting, exposure and other tricks.
Real world test
I’ve created a fairly accurate real world test to make sure I’m on the right track. First I scanned the Wolf Faust Target I mentioned earlier and brought it into PhotoShop applying the color profile I created and converted it to Adobe RGB. Without adjusting the image I plotted it and compared the print with the actual target. My test with the HP 5500 large format printer on HP satin photo paper proved to be amazingly accurate.
Next, in PhotoShop, I created texture bitmaps from the scanned target. In this test, I created nine bitmaps, one for Red, Green, Blue, Yellow, Magenta, Cyan, White, Gray and Black. I mapped these onto cubes in a simple room that has a neutral color (value 145) and white overhead lights. I then created materials using the actual LAB numbers that were used to create those colors on the target (translated to RGB) onto a second set of cubes in the room and rendered. Finally, I brought the rendered image into PhotoShop, assigning Adobe RGB and printed. In comparing the original target to the final print the colors were, in my opinion, very close and I feel confident that my process is accurate.
The image below shows the final rendered output with LAB (RGB) colors and scanned texture maps mapped to cubes and rendered. Also shown are the bitmap images and the LAB colors used in the rendering. In the middle are the white, gray and black maps and colors. Notice the huge discrepancy in the gray color (right) vs. the gray map (left). I’m still trying to figure that one out. But the other colors look remarkably similar. Since adopting this workflow, I’ve found that my renderings have become much more accurate.
Here is what a color target looks like.
The intent of this article is to point out that a linear workflow is not always necessary nor is it always the best solution. I know that there are many people whose work requires a different approach than that which I outline below. But since I’ve had great success with this process, I thought I’d share it in hopes that there are some who will benefit by it.
My work consists mostly of creating architectural presentations shown to a non-architect audience. My mandate is to produce the most photorealistic images possible. The initial presentations are projected electronically whereas the final presentation consists of 32 x 40 inch prints. I use 3D Studio Viz, Vray 1.47.03 and PhotoShop CS2 on WinXP pro.
Because of my workflow requirements, I have decided that using gamma 2.2 is better for me than 1.0 or linear. There are several reasons for this. First, most of the many hundreds of monitors in our organization are not calibrated. Those that are have been calibrated to gamma 2.2. Of course, I could always apply the appropriate profile before distributing my images, but I don’t like having to keep track of different versions of my files. Plus the process adds another step that takes time.
Second, most monitors perform much better at 2.2 than at 1.0 and gamma 2.2 images are displayed more accurately, even though the file itself might be limited due to the gamma curve. Also, calibrating a Windows PC system to 1.0 makes the desk top so washed out it is really uncomfortable to use, even when using dark desktop colors.
Third, most of our peripheral devices such as projectors, printers, scanners and plotters are intended to output 2.2 images. In tests I’ve performed, I’ve found that there is no visual difference in output quality between linear images vs. gamma 2.2 images. Even the PhotoShop tests described on the AIM website http://www.aim-dtp.net./aim/evaluation/gie/index.htm don’t seem to have as big an impact as claimed. In short, the added steps needed to deal with 1.0 images in my workflow just aren’t justified.
While I understand the logic behind a linear workflow, I’ve found that using a consistent 2.2 process together with good color profile management and proper monitor calibration makes up for any reduction in file integrity caused by the 2.2 gamma.
Workflow
My work typically begins with architectural massing and volume models that I render using monochromatic colors, usually just tonal variations of gray and white. I use values around the 130 – 140 range. Accurate lighting is important at this stage as it helps define the architectural forms and gives depth. For interiors I use slightly cool plane lights at window openings and slightly warm sphere lights at fixtures, both with a multiplier of 1 ~ 2. For exterior work, I use a target direct light with a slightly warm color, multiplier 1, and a slightly cool environment skylight, multiplier .35. The second stage involves importing scanned images as texture maps, tweaking the lighting, adding furnishings and fixtures for interior images and landscape and entourage for exterior images. The third step is to make final lighting and material tweaks and create the final renderings. The final step is to bring the image into PhotoShop, tweak colors once more, add a few blending layers for softening effects and print.
Here’s how I set up my system and software:
Monitor calibration
Many hardware devices are available and people in our organization use them frequently. However, I’ve gotten virtually the same results using Adobe Gamma and the calibration image and procedures found here:
http://www.aim-dtp.net./aim/calibration/index.htm
and here:
http://www.aim-dtp.net./aim/calibrat...d_contrast.htm
Scanner calibration
I use a flatbed scanner for most of my texture capturing but occasionally use a digital camera. In either case the input device must be properly set up and an accurate profile assigned. The profile is created by comparing a scan or digital photograph of a Color Target to the LAB colors that were used to create the target. Any discrepancy is recorded and a formula or Profile created to correct for those differences. Information can be found here:
http://www.coloraid.de/
In my scanner setup, I select output to gamma 2.2. Comparisons between 2.2, 1.0 and non corrected scans favor 2.2. Of course, this might be different for other scanners.
PhotoShop Color Management
The key to the whole process is using appropriate color settings and profiles. In the color settings dialog you will see a whole bunch of choices. I know there is debate on which settings to use, but my preference is to use Adobe RGB (199
for the working RGB space and US Web Coated (SWOP) v2 for the CMYK space. I prefer Adobe RGB because it has a gamut that is more encompassing than the smaller gamuts like AppleRGB, ColorMatch RGB or sRGB and is more closely aligned to the NTSC gamut. As for the CMYK color space, the SWOP v2 space gives me slightly better print results on my HP 5500 plotter than using the Adobe RGB setting. Since I allow my Printer to Determine Colors (selected in the print dialog box) I make sure that Color Management is turned on in the printer driver and that the printer firmware is selected to the appropriate CMYK profile.The other Color Settings in PhotoShop are as follows: Mismatched and Missing Profiles: set all to Ask When Opening. Engine: Adobe (ACE). Intent: Relative Colorimetric. Use Black Point and Dither: both checked. Desaturate and Blend: unchecked.
3D Studio Settings
In Preference Settings I check Enable Gamma Correction and set my Display and Bitmap Gamma all to 2.2. If I get a bitmap file that is different than 2.2, I override it in the load bitmap dialog box.
For color mapping, I use Linear Multiply, Dark1.0, Light1.0 & Clamp Output unchecked. I’ve tested different settings and these work best for my workflow. Using Gamma Correction Dark1.0, Light .455 and saving the file with gamma 1.0 comes close but textures come out a bit too dark, colors over-saturated and for some reason the AA is not as good. (To compensate for the 1.0 output, I assigned the AIM RGB Trinitron D65 G1.0 profile. Assigning other 1.0 profiles had a similar effect).
Practical application of the settings
Since most of my textures and materials exist in the real world, I can compare the scanned image and the rendered 3D image with the actual physical object. This provides a good test to the accuracy of the calibrations and settings.
When I open a scanned file or digital image, PhotoShop asks what I want to do. Since I created a scanner profile, I select Assign Profile then select the profile that I created and check Convert Document to Working RGB. Since my monitor is calibrated accurately and I’ve created a good scanner profile, what I see on my screen should look very close to the actual object. When I save it, I make sure that the Adobe RGB profile is saved with it by checking the ICC Profile check box in the Color section of the Save As dialog box.
When I import the texture map into 3D Studio, I make sure it looks the same in the Material Editor as it does in real life, another good check for accuracy. When I render the scene, I try to set up my lights so that the rendered material looks close to the real object in tone and color. Only after I’m satisfied with the exposure and color of the texture do I adjust the lighting for mood and effect. When I save the rendered file and load it into PhotoShop, I again compare the result with the actual materials that I scanned to make sure I’m not faking things too much with lighting, exposure and other tricks.
Real world test
I’ve created a fairly accurate real world test to make sure I’m on the right track. First I scanned the Wolf Faust Target I mentioned earlier and brought it into PhotoShop applying the color profile I created and converted it to Adobe RGB. Without adjusting the image I plotted it and compared the print with the actual target. My test with the HP 5500 large format printer on HP satin photo paper proved to be amazingly accurate.
Next, in PhotoShop, I created texture bitmaps from the scanned target. In this test, I created nine bitmaps, one for Red, Green, Blue, Yellow, Magenta, Cyan, White, Gray and Black. I mapped these onto cubes in a simple room that has a neutral color (value 145) and white overhead lights. I then created materials using the actual LAB numbers that were used to create those colors on the target (translated to RGB) onto a second set of cubes in the room and rendered. Finally, I brought the rendered image into PhotoShop, assigning Adobe RGB and printed. In comparing the original target to the final print the colors were, in my opinion, very close and I feel confident that my process is accurate.
The image below shows the final rendered output with LAB (RGB) colors and scanned texture maps mapped to cubes and rendered. Also shown are the bitmap images and the LAB colors used in the rendering. In the middle are the white, gray and black maps and colors. Notice the huge discrepancy in the gray color (right) vs. the gray map (left). I’m still trying to figure that one out. But the other colors look remarkably similar. Since adopting this workflow, I’ve found that my renderings have become much more accurate.
Here is what a color target looks like.
Fact is I work completely on non-disclosed projects which is why I've never posted before. BBullo, back me up on this one. But I'll see if I can persuade my boss to make an exception. If not, I'll just have to find an old project from my glory days, render and post it.
Comment