Research And Implementation Of Image Denoising Based Android

Recently, phones based Android has become commonplace in people’s lives. Using a Android phone to take photos can be seen everywhere, especially in the promotion of social software like qq and weixin. Self-timer has become one of the most popular acts at the moment. So the image quality is good or bad, has become a important indicator of performance to measure a mobile phone. A android phone carema is made by COMS(Complementary Metal Oxide Semiconductor), which make the camera carrying natural transient noise, similar to Gaussian noise. Manufacturers usually improve the quilty of a picture by increasing the rear camera pixels or increasing the amount of exposure, etc. But the front camera is lower than the bank, the images always contain a lot of noise under conditions of lower light。The performance of phones with different Hardware are always not alike. Anroid phones with a GPU have much higher floating-point operations per second than that not have one. Image denosing is very populor in many area, Selecting the approprite denoising algorithm according the phone’s hardware is extremely important considering the user experience. This paper do some research about denoising algorithm based Android phone, Its main point has four parts.1. Wavelet threshold denoising. The time complexity of Wavelet is relatively low,and its effect not bad which met the requirement of the phone.this paper will discuss several common wavelet denoising algorithm based their simulation. Select the BayesShrink threshold at last.2. Non-local means denoising. Study basic NLM denoising algorithm, analyzing the original NLM denoising algorithm problems within effect and time complexity. Using the CPWS, combining James-Stein shrinkage to improve the effect, using the FFT and SSI to reduce time complexity. Analysis the FLJS-NLM based the simulation.3.Implementation of BayesShrink threshold with Android. Analysis the specific progress of Android camera based Android system architecture. Perform the BayesShrink thresholding algorithm by using JNI and NDK programming techniques, generate dynamic link library, complete denoising software based phone CPU.4. Implementation of FLJS-NLM with Android. Study heterogeneous computing on GPU platforms, and programming FLJS-NLM algorithm in android system by using OpenCL. complete denoising software based phone GPU.