A process for evaluating the quality of coded images (e.g., two video sequences) uses a homogeneous motion vector field and psychovisual filtering to take into account Human Visual System (HVS) characteristics to calculate the error between the two video sequences.
The present invention provides a quantization control system for video coding, for optimizing bit allocation based on visual priority which expresses at high precision human visual characteristics for videos, thereby improving subjective image quality of the whole image even with a limited number of bits. An object is extracted from a picture of an input image (S1) and a gaze parameter V(j) for the extracted object is calculated (S2). Then, a texture attribute parameter t(k) of each macroblock included in the object is calculated (S3). The gaze parameter V(j) and texture attribute parameter t(k) are used to calculate a visual priority parameter w(k) for each macroblock. Based on this parameter w(k), a quantization parameter is determined.
An image processing apparatus changes a pixel value of a pixel of interest in image data to be processed by a predetermined condition concerning a spatial frequency to generate corrected image data, and performs a predetermined coding process with regard to the corrected image data to generate coded data. Thus, processor load in the coding process can be reduced.
