Apparatus and method for combining random set of video features in a non-linear scheme to best describe perceptual quality of video sequences using heuristic search methodology

摘要

In order to assess the perceptual quality of the video sequence, a method for combining a set of the feature video random non-linear is,; (A) receiving a video sequence to the image quality;; (b) cross-correlation information, the method comprising: providing an objective image-quality metric comprises a controller for a random set of metric of M ₁ to M _n range without (cross correlatoin information);; (c) the metric of the respective one of the metrics of the set of metrics to provide the random one of the metric x ₁ to x _n individual objective scoring value (scoring value) of the video sequence in the range of each set individually applying the video sequence;; (d) determining the weights (weights) a plurality of sets of (w ₁ to w _n) for correlating a predetermined number of video sequences (n) a predetermined subjective evaluation of image quality for the by-weight number of each set of the set of weights has the crystal, which is assigned to the range having the incremented value, such as the range is divided by the number of combinations for each of the set of weights step;; (e) step (c) the weight of an objective scoring values x ₁ to x _n for each individual as a single set of each of the weights being provided by one of the metrics of said each of said random set of metrics in .;; (F) the objective scoring values ​​of the individual weights of the random a set of metrics, the method comprising: adding to a single objective evaluation of F, a weight of each of the individual scoring values ​​from step (e) are respectively from step (c) objective scoring value for each x ₁ to the addition step is multiplied by x _n;; (G) calculating a correlation factor R in order to provide a correlation value for the objective evaluation of F and a number of the video sequence (n);; (H) repeating the multiple correlation factor of step (e), (f) and (g) for each of the set of weights provided in step (d) to determine the R;; (I) the method comprising the s R Any number of the acquired ranking (ranking), a specific correlation factor of the closest a particular correlation value multiple correlation factor the having the first step for each of the set of weights (e ), the ranking step in displaying the best ranking of each of the combined metric;; (J) the best perceived image quality (perceptual image quality) for providing to each of the system optimizer for optimal ranking of the combined metric image quality to at least one of the (system optimizer) and video processing module obtained in step (i) to and a step of providing the information. Step (e), (f) and (g) are repeated to provide a plurality of correlation factors that rank. Search perceptual uses a genetic algorithm to find the optimal set of weights to provide an objective score closest to a predetermined subjective evaluation. The system provides the hardware and module to run a non-linear combination of metrics to provide an improved perception of the image information.