Image compression method, device and system and storage medium

Abstract

The invention provides an image compression method, device and system and a storage medium. The method comprises the steps: performing scene recognition on a to-be-processed input image, wherein the scene recognition comprises calculating a brightness value of the input image; determining a compression curve corresponding to the scene recognition result based on a preset compression curve corresponding to the preset brightness value; and compressing the input image based on the determined compression curve corresponding to the result of the scene recognition. According to the image compressionmethod, device and system provided by the embodiment of the invention, a plurality of compression curves are predefined to correspond to optimal compression curves under different scenes (brightness); therefore, the compression curve of the input image is adaptively generated according to the scene, the color of the input image in the full-brightness domain can be accurately restored, and the color information and contrast of the input image are maintained.

Description

technical field [0001] The present invention relates to the technical field of image processing, and more particularly relates to an image compression method, device, system and storage medium. Background technique [0002] As camera equipment becomes more and more popular, people's demand for photography is getting higher and higher. Shooting high dynamic range scenes is one of the more difficult scenes. The reasons include: 1. Due to the large dynamic range in the scene, it is necessary to store images through high bit width. The source of high bit width images includes high-end sensors or multiple low bit width images. High dynamic range stitching of images, and these operations need to be performed on the original (RAW) image, and the RAW image only has information of one of the three channels of RGB on each pixel; 2. The RAW image needs to pass through the image signal processor (ISP) can become a real digital image after processing, and ISP is extremely sensitive to b...

AI Technical Summary

Problems solved by technology

Among them, linear compression considers the distribution of pixel colors, which will reduce the color resolution in the numerical value set, and the mapped image will cause the problem of loss of details in dark areas.

Logarithmic compression cannot effectively correct the local part of the image, which will lead to the problem of reduced contrast in bright areas

Piecewise linear compression may cause contrast or color cast problems in linearly mismatched color ranges

Tone mapping compression is a non-global image operation, which will cause difficulties in subsequent image processing, resulting in color cast or contrast problems in the final output image

Images