A novel image compression method for use in plant pathology diagnosis for making use of the networks

摘要

To reduce the prevalence of illness and detect physiological disorders in their early stages, pathological diagnosis of plants is important. When a plant image stored as a large data file is sent, transmission takes a long time. To reduce transmission time, image compression is needed. The most widely used compression format for photographs is JPEG (Joint Photographic Experts Group). The advantage of JPEG is its high compression rate. A weak point is irreversibility; it is impossible to reproduce the image perfectly.We use plant images for pathology diagnosis and investigated an algorithm to compress such images without loss of information. We adopted a method that separates the photographic image into useful parts (plant) and other parts (background), and stores and compresses only the useful part. The algorithm is as follows: 1. Mask to get the plant information for diagnosis from a photographic image; 2. Delete the background; 3. Compress and tore the image of only the plant image. We wrote a program to recognize and to mask the plants based on HIS (Hue, Saturation, Intensity). The program covers not only green vegetables, but also fruits. Furthermore, the program was investigated in combination with existing image compression techniques (for example, tiff and tga). We achieved a compression rate of about 45 percent compared with original photographs. We expect the compression rate to be increased in the future by optimizing the HIS parameters.