Palette device and generation method for image with transparency information

Abstract

The invention discloses a palette device and a generation method for an image with transparency information. The method comprises the following steps: defining the number of independent entries in an appointed palette as K; quantifying RGB (Red, Green and Blue) three primary colors and transparency value of the image with the transparency information; generating K RGB palette entries with Alpha transparency value; obtaining index values of all pixels of the image; and finally saving the values to form a PNG (Portable Network Graphic) 8 image. According to the invention, the RGB three primary colors and the Alpha transparency are quantified; during quantification, taking the influence of the Alpha transparency of the image to the display effect of the RGB three primary colors into consideration, the quantification series at a lower Alpha transparency value are reduced, and the palette entries are subjectively distributed in a more reasonable manner; when the length of the palette is smaller, a palette initial value obtained according to pixel frequency is subjected to LBG vector quantification, so as to avoid the phenomenon that too much colours lose due to small frequency and effectively reduce the distortion of the quantified image; and when the length of the palette is larger, a Freud jitter method is adopted in the image quantification process so as to effectively improve the subjective quality of the image.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an image palette device and generation method with transparency information. Background technique [0002] A commonly used color image in RGB888 format can contain up to 2 24 kinds of colors. With the continuous development of computer graphics and image technology, the transparency / translucency of the image is added to the RGB three-color basis to make the image display more flexible. For example: in the PNG image format developed in the mid-1990s, its PNG32 image adopts the RGBα8888 format, that is, each channel of R, G, B, and α uses 8bit quantization respectively. PNG stores color images with a depth of up to 48 bits, and can also store alpha channel data up to 16 bits. In other words, the maximum number of colors that a PNG image with alpha channel can contain is far greater than 2 24 . [0003] On the one hand, image display devices can only display a limited number of ...

AI Technical Summary

Problems solved by technology

[0008] (1) The MCQ algorithm has a large memory overhead and requires enough memory as a counter for each color in the source image

In addition, the local details of the image reconstructed by this method are relatively blurred, and the effect is not ideal.

[0009] (2) In the octree algorithm, in the process of simplifying the octree, the combination of colors is based on the frequency of pixels, so some colors will still be lost due to the small frequency of occurrence, resulting in image display distortion

[0010] (3) The above algorithm only supports the quantization of RGB or other non-transparency colors

Although the author of the median segmentation algorithm proposes that the algorithm can be extended to more channels, the smaller the value of the α channel, the lower the visual sensitivity to image color. If only the α channel is added to the median segmentation quantization as an independent channel, then The amount of calculation and the system delay increase proportionally, which increases the processing cost

Images