Image quality (often Image Quality Assessment, IQA ) is an image characteristic that measures perceived image degradation (typically, compared to ideal or perfect images). The imaging system may introduce some amount of distortion or artifacts in the signal - for example by transcoding -, which affects the quality experienced subjectively and the Quality of Experience for the end user.
Video Image quality
In photographic imagery
A (picture) is formed on the camera image area and then measured electronically or chemically to produce the photo. The process of image formation can be explained by the ideal pinhole camera model , where only the light rays of the illustrated scene pass through the camera aperture can fall on the image plane. In fact, this ideal model is only an approximate image formation process, and the image quality can be explained in terms of how well the camera approaches the pinhole model.
The ideal model of how a camera measures light is that the resulting image must represent the amount of light that falls at any point at a given point in time. This model is only an approximate description of the camera's light measurement process, and image quality is also associated with deviations from this model.
In some cases, the image whose quality should be determined is primarily not the result of the photographic process in the camera, but the result of storing or transmitting images. Common examples are digital images that have been compressed, stored or sent, and then decompressed again. Unless a lossless compression method has been used, the resulting image is usually not identical to the original image and the deviation from the original image (ideal) is then a quality measure. Taking into account a large number of images, and determining the quality measures for each of them, statistical methods can be used to determine the overall quality measure of the compression method.
In ordinary digital cameras, the resulting image quality depends on the three factors mentioned above: how much of the camera image formation process deviates from the pinhole model, the quality of the image measurement process, and the encoding artifacts introduced in the image produced by the camera, usually by the method JPEG encoding.
By defining image quality in terms of deviations from ideal situations, quality measures become technical in the sense that they can be objectively determined in terms of deviations from the ideal model. The quality of the image can, however, also be related to the subjective perception of an image, for example, a human looking at a photograph. An example is how color is represented in black and white images, as well as in color images, or that the image quality reduction of noise depends on how the noise correlates with the information the viewer seeks in the image rather than the whole. power. Another example of this type of quality measure is Johnson's criterion for determining the image quality needed to detect targets in the night vision system.
Quality subjective measurement also relates to the fact that, although camera deviations from the ideal model of image formation and measurement are generally undesirable and corresponding to the deterioration in the quality of objective images, these deviations can also be used for artistic effects in image production, according to high subjective quality.
Maps Image quality
Image quality rating categories
There are several techniques and metrics that can be measured objectively and automatically evaluated by computer programs. They can be classified depending on the availability of reference images or features of the reference image:
- Full reference method (FR) - FR metric trying to judge the quality of test images by comparing them with reference images assumed to be of perfect quality, e.g. original image versus JPEG compressed image version.
- Re-reference method (RR) - RR metrics assess test quality and reference images based on feature comparison extracted from both images.
- Unassigned method (NR) - NR metrics try to judge test image quality without any reference to the original.
Image quality metrics can also be classified in the size of only one specific type of degradation (for example, obscuring, blocking, or ringing), or considering all possible signal distortions, ie some types of artifacts.
Image quality metrics
A large number of image quality models have been developed in the last decade, mostly derived from academic research.
Full reference metric
Well-known and frequently used metrics include Peak Signal to Noise Ratio (PSNR), Structural Equalization (SSIM), and Visual Fidelity Information (VIF).
Most image-quality measuring tools commercially deployed by television and home cinema industry use SSIM. VIF serves as a core image quality prediction engine in VMAF Netflix video quality monitoring system, which controls all the Netflix encoded videos that are streamed around the world. Both SSIM and VIF are developed in the Lab for Images & amp; Video Engineering (LIFE).
Metrics Without reference
Source of the article : Wikipedia
0 Comments