Perceptible Variance between RAW and JPEG Images

To detect any perceptible variance between processed RAW and JPEG Images, a test was conducted. In a series of frames, the test subject was underexposed one, two, and three exposure stops, and then overexposed one exposure stop, in both the RAW (.NEF) and JPEG file formats.

During processing, the requisite exposure value necessary to bring each image back to correct exposure was made. For example, if an image was exposed -1 EV during capture, then it was processed at +1 EV. The RAW files were processed in Adobe Camera RAW (ACR), and the JPEG images were processed in Photoshop.

A Nikon D70 camera, and a Nikkor AF 20-35mm EDIF F/2.8 lens (at 35mm) were used for the test. The aperture was f/16 and the ISO rating was 400 on all shots. The images were captured in Aperture Priority mode, with Matrix metering. The camera and lens were mounted on a Gitzo tripod, and the lighting on the test subject was natural directional lighting from the left side. The JPEGs were taken at the largest file size (same size as the .NEFs) with the detail setting at "normal".

The only variable in the compared images at any given exposure value (EV) was the format: one was RAW (.NEF) and the other was JPEG. After the test images were processed, they were viewed in Photoshop at 100%. Four different areas of the subject were sampled and compared. The table below, Figure 1, reveals the increase or decrease in exposure value (EV) each sample received during capture and during processing.

Sample	Capture	Proccessing
1	-1 EV RAW	+1 EV RAW
2	-1 EV JPG	+1 EV JPG
3	-1 EV RAW S	+1 EV RAW S
4	-1 EV JPG S	+1 EV JPG S
5	-2 EV RAW	+2 EV RAW
6	-2 EV JPG	+2 EV JPG
7	-2 EV RAW S	+2 EV RAW S
8	-2 EV JPG S	+2 EV JPG S
9	-3 EV RAW	+3 EV RAW
10	-3 EV JPG	+3 EV JPG
11	-3 EV RAW S	+3 EV RAW S
12	-3 EV JPG S	+3 EV JPG S
13	+1 EV RAW A	-1 EV RAW A
14	+1 EV JPG A	-1 EV JPG A
15	+1 EV RAW B	-1 EV RAW B
16	+1 EV JPG B	-1 EV JPG B

Figure 1. Exposure Values

The test subject was the Quaker Oatmeal box in Figure 2, below. The samples titled +1, +2, and +3 EV were selected for midtone comparison. Samples +1, +2, and +3 EV S were selected for shadow comparison. And, samples -1 EV A and -1 EV B were selected for highlight comparison.

Test Subject, Quaker Oatmeal Box

Figure 2. Test Subject

The Results: Samples of Test Images at 100%

Revealing Exposure Adjustment During Processing

(+1 EV RAW) Sample 1	(+1 EV JPG) Sample 2


(+1 EV RAW S) Sample 3	(+1 EV JPG S) Sample 4


(+2 EV RAW) Sample 5	(+2 EV JPG) Sample 6


(+2 EV RAW S) Sample 7	(+2 EV JPG S) Sample 8


(+3 EV RAW) Sample 9	(+3 EV JPG) Sample 10


(+3 EV RAW S) Sample 11	(+3 EV JPG S) Sample 12


(-1 EV RAW A) Sample 13	(-1 EV JPG A) Sample 14


(-1 EV RAW B) Sample 15	(-1 EV JPG B) Sample 16

Observations

1. The processed underexposed RAW and JPEG files each reveal different levels of noise directly proportional to the degree of their respective underexposure. However, at any given level of underexposure, the noise was more noticeable and pronounced in the 8 bit JPEG files, than it was in the 12 bit RAW files.

2. In both formats, at any given level of underexposure, noise was more noticeable and pronounced in the shadows, than it was elsewhere in the image.

3. When the 8 bit JPEG test image was overexposed +1 EV, lost highlights could not be recovered during processing. See, Samples 14, 16. When the 12 bit RAW test image was overexposed +1 EV, at least some of the highlights were recovered during RAW conversion. See, Samples 13,15.

4. The latitude for correcting overexposure and underexposure while processing RAW files is better than with JPEGs, likely due to the presence of more tonal values in the 12 bit files.

< back to article >