JPEG and Gamma Compensated Images
Some mis-information of Mr. Poynton clarified.
In section 19. "Should I do image processing operations on linear or nonlinear image data?" of his so called GammaFAQ.html Mr. Poynton claims: "... The image processing literature rarely discriminates between linear and nonlinear coding. In the JPEG and MPEG standards there is no mention of transfer function, but nonlinear (video-like) coding is implicit: unacceptable results are obtained when JPEG or MPEG are applied to linear-light data. ..."The below experiment was carried out in order to see if the above statement holds.
Please download the original in case you wish to re-create these experiments.| On this left column the linear-light original was converted to JPEG, by first saving the image as JPG. This image was then gamma compensated for viewing on uncalibrated PC Monitor. Then the image was saved again using the same JPEG compression settings. | On this right column the linear-light original was first gamma compensated for viewing on uncalibrated PC Monitor and then converted to JPEG, by saving the image as JPG. |
| Note: To apply the JPEG over a linear-light image and then to show this image on uncalibrated systems requires two lossy JPEG Save operations. First the linear image is saved as JPG, then it is gamma compensated and then saved again. This is normally never done, but it is the only way to achieve the condition that is described in the GammaFAQ of Mr. Poynton. | The
gamma compensated images suffer only one lossy save operation. The
linear-light original is first gamma compensated and then saved as
JPG.
This is how saving to JPEG is normally done. |
2. Saved as jpg at level 15. 3. Gamma compensated by 2.5 4. Saved again. |
2. Gamma compensated by 2.5 3. Saved as jpg at level 15. |
2. Saved as jpg at level 4. 3. Gamma compensated by 2.5. 4. Saved again. |
2. Gamma compensated by 2.5. 3. Saved as jpg at level 4. |
2. Saved as jpg at level 10 (max quality). 3. Gamma compensated by 2.5. 4. Saved again. |
2. Gamma compensated by 2.5. 3. Saved as jpg at level 10 (max quality). |
Conclusions
Mr. Poynton's claim in the GammaFAQ is false. If there is a difference between the images it is in favor to the linear-space editing despite the fact that they had experienced two lossy save operations. Note: Gamma compensation in Photoshop was done using an accurate inverse gamma 2.5.amp file in the curves dialog.Accurate Image Manipulation for Desktop Publishing