For the defect in describing affine and blur invariable of scale-invariant feature transform (SIFT) at large viewpoint variation, a new object recognition method is proposed in this paper IntroductionIn the object recognition with complicated background or occlusion, local feature is better than global feature in stability, repeatability and authenticability and it has been widely applied in image matching, machine vision and other fields in recent years. This paper mainly makes in-depth research to the detection and description of local region feature.Scale-invariant feature transform (SIFT) [1] algorithm has excellent scale invariance and rotation invariance in feature point extraction in linear scale space and the main direction of local gradient distribution, but it has no affine invariance. Compared with blotch feature, the region detection methods proposed in recent years are applicable to the feature region detection of various shapes and it can preserve excellent invariance even when the view-angle changes greatly. Literature [2] has made comparative analysis in such methods as SIFT, HarrisAffine, Hessian-Affine and maximally stable extremal region (MSER) [3] region detection which is proposed by Matas and the result shows that MSER has the best detection effects in recognizing gray-level consistency region with strong boundaries to be recognized, view-angle changes and light variation; that when the image scale changes, MSER follows only after Hessian-Affine and that when the image is fuzzy, MSER is the most non-ideal in performance. The research result of Literature [4] shows that SIFT has better description effect in plane objects, but MSER has excellent description effect in most natural scenes.In the local feature description, plenty of local feature descriptors have been proposed in recent years and their performances are significantly different in different applications; however, there is no universal description algorithm. Literature [5] and [6] analyze the performance of the local feature descriptors which are proposed in the past years from different perspectives and the analytical result demonstrates that the SIFT descriptor based on one-order histogram has the best scale invariance and MROGH [7] has the best performance in light variation. Huang and others have come up with the local feature descriptor based on the distribution of the histograms of second-order gradients (HSOG) and it excels in describing the geometrical features related to curvature; however, it is low in the recognition efficiency of second-order histogram. Therefore, this paper integrates MSER detection and SIFT description and uses the improved MSER to detect the local objective local feature region and SIFT to construct feature descriptor for object recognition.

Due to the limitation of imaging equipment, the influence of transmission medium and external environment, image quality degradation will inevitably occur in the process of generation, transmission and reception. These degradation not only worsens the visual effect of the image, but also makes the image lose a lot of useful information, which seriously affects image recognition, target detection and other high-level visual analysis. Wavelet analysis can extract useful information from image signal and meanwhile its profound wavelet basis can get adapted to signals of different properties. To better apply wavelet transform into image restoration domain, this paper according to the characteristics of wavelet transform, analyzes the method to select threshold function and the relationship within and between layers of wavelet coefficients, gets a proper threshold weight coefficient and propose an adaptive weighted threshold image restoration method based on wavelet domain, which makes smaller deviation and variance between the de-noised image and the original signal. The experiment result shows that the algorithm of this paper can obtain good subjective and objective image quality and effectively retain most detailed information of the image.

With the advancements of computer technology, image recognition technology has been more and more widely applied and feature extraction is a core problem of image recognition. In study, image recognition classifies the processed image and identifies the category it belongs to. By selecting the feature to be extracted, it measures the necessary parameters and classifies according to the result. For better recognition, it needs to conduct structural analysis and image description of the entire image and enhance image understanding through multi-object structural relationship. The essence of Radon transform is to reconstruct the original N-dimensional image in N-dimensional space according to the N-1 dimensional projection data of N-dimensional image in different directions. The Radon transform of image is to extract the feature in the transform domain and map the image space to the parameter space. This paper study the inverse problem of Radon transform of the upper semicircular curve with compact support and continuous in the support. When the center and radius of a circular curve change in a certain range, the inversion problem is unique when the Radon transform along the upper semicircle curve is known. In order to further improve the robustness and discrimination of the features extracted, given the image translation or proportional scaling and the removal of impact caused by translation and proportion, this paper has proposed an image similarity invariant feature extraction method based on Radon transform, constructed Radon moment invariant and shown the description capacity of shape feature extraction method on shape feature by getting intra-class ratio. The experiment result has shown that the method of this paper has overcome the flaws of cracks, overlapping, fuzziness and fake edges which exist when extracting features alone, it can accurately extract the corners of the digital image and has good robustness to noise. It has effectively improved the accuracy and continuity of complex image feature extraction.