Image Feature Coding And Its Application

With the rapid development of image classification,saliency object-based detection,and person re-identification,the demand of automatical image analysis becomes more and more urgent in real environments.Saliency object detection is to detect salient object of scene images,which is an important mechanism of people vision system.Image classification is an important aspect for image understanding,and its task is to classify image to different semantic images.Person re-identification is increasingly playing an important role in people's life,which refers to recognizing the same person who moves cross the multi-shots.In this dissertation,we analyze image from three aspects,that is image classification,saliency object-based detection,and person re-identification.The main contributions of this dissertation consist of the following parts.1.Different from the usual way of concatenating or merging the category codebooks to form a global dictionary,we employ the category codebooks to calculate a type of categorysensitive saliency feature,and then,encode the saliency features to form a representative of image content.Thus,we present a feature encoding scheme for image classification by combining the salient coding method with the category-specific codebooks,which are generated separately using the training images of each category.Compared to the state-of-the-art methods such as LC-KSVD,the dictionary generation and feature encoding in our scheme are pretty simple,and no complicated optimization is involved.However,our scheme can achieve better,in some cases,significantly better results,in terms of the classification accuracy,than the state-of-the-art methods.The accuracy of image classification is 79.8%.2.Predicting object location using a top-down saliency model has grown increasingly popular in recent years.In this work,we combine Locality-constrained Linear Coding(LLC)with a Conditional Random Field(CRF),and construct a top-down saliency model to generate a specific object-based saliency map.During the training phase,we use the LLC codes as the latent variables of the CRF model,and meanwhile learn a class-specific codebook by CRF modulation.In the testing phase,we use this top-down model to distinguish specific objects from a cluttered background.Finally,we evaluate the experimental results on the MSRA-B,Garz-02,Weizmann Horse,and Plane datasets by applying the developed object-based saliency model.The performance shows that our approach can not only improve the precision but also dramatically reduce the computational complexity.3.Computing the salient object region in real scenes has gained significant progresses in recent years.In this work,we propose a novel method for computing saliency object region by boosting background information and top-down visual saliency model,which is well-suited to locate category-specific saliency objects in cluttered real scenes.First,we use robust background measure to acquire clean saliency maps by optimizing background information.Second,we learn a top-down saliency object model by jointly class-specific codebook and a Conditional Random Field(CRF)during training phase.Furthermore,the key point of our model is to use the Locality Linear Codes as the latent variables of CRF.Finally,we compute the saliency object region by boosting robust background measure and top-down model.Experimental results on the Graz-02 and PASCAL VOC2007 datasets show that our method is able to achieve much better saliency object maps than the state-of-the-art.4.In this work,technologies of image processing and computer vision is used to solve the person re-identification problem: As for pedestrian detection,HOG and SVM is employed to find pedestrian in videos;In terms of feature extraction,an image is described by color histogram and texture with the help of image segmentation,kernel function and spatial pyramid;In case of classification and matching,an effective feature coding algorithm is implemented to build the corresponding relations between probe image and gallery set.In practical application,realtime response is as important as re-identification rate.Nonetheless,most person re-identification researches care less about computation time cost in their work.This paper proposes an algorithm that not only achieves more accurate re-identification rate but also reduces computation complex.