Research Of The Algae Image Texture Based On Morphology And None Entirely Tree Wavelet Decomposition

The algae are most important not only as primary producers but the resources givers in the marine ecosystem. They play a critical part in material cycling and emery flowing. However, the booming of certain algae could bring about natural disasters, which can damage the marine ecosystem. Therefore, algae supervision and identification are much more concerned.The mud and unknown objects are inevitably brought into the imaging procedure, which cause great noises. What's more, the edges in these images are not clear and continuous as a result of the equipments. Thus, the preprocessment of images is very critical. This article is based on the mathematics of the image preprocessing method, using such algorithms like control mark operator and high hats—low caps to denies. At the mean time, the OTSU is conducted to maintain details while extracting the useful information.Many researches have done in the wavelte analysis and the texture features extraction. Finally, by the use of none entirely tree wavelet decomposition, direction and the number of details are chosen as the characteristics for identification. Unlike the traditional wavelet decomposition, a threshold is used to decide which sublayer should continue the next decomposition. This is helpful both in mainting the details and in compressing the characteristics of information. Also, the distinction method of consistency is proposed in the wavelet decompositions to further improve the charactersis extraction speed.DPNN is used to identify images in this paper. Firstly, Brodatz database is simulated to prove the efficiency and speed accerlation of this algorithm. And then the characteristis : numbers of details and directions, are sent to DPNN for identification. According to characteristics of algae images collected at scene, researches have done in prepocessment, texture extraction and characteristic identification. During preprocessing, a combination of high caps– low caps and OTSU is proposed to denoise and enhance the details; SOBEL is used to extract the edge. In the form of identification, the distinction method of consistency is proposed in the wavelet decompositions to further improve the charactersis extraction speed.