"Image Retrieval By using Visual Descriptor"

The Java Application System which retrieve similar Images from large database after providing the query image by the user using Image retrieval by using Visual descriptor on Hadoop.

A Block Diagram of Image Retrieval

Literature Survey:

Image retrieval system is concerned with searching and retrieving of digital images from collection of large digital image database. The uses of text description to the images or visual features like color, shape, texture have been used for image retrieval from database. Different types of image retrieval techniques are discussed below

1. Text Based Image Retrieval

2.   Content Based Image Retrieval

1.Text Based Image Retrieval:

In early 1970s, text based image retrieval was widely used framework of image retrieval to annotate the images by text to perform image retrieval. In TBIR, the images are manually annotated by text descriptors.

Text descriptors are sometimes inaccurate due to the subjectivity of human perception on complicated image feature very well. This technique requires substantial level of human labor for manual annotation which is impractical for very large databases. To overcome all these drawbacks content based image retrieval is introduced.

2. Content Based Image Retrieval:

• Content based image retrieval is an application of computer vision technique to address the problems allied with text based image retrieval in large digital image database. In content based image retrieval the query is in the form of image and its low level features are used as the content describing it.
•  Low level features are set of characteristics of the image such as color, texture, and shapes. These features are extracted from the query image as well as for all the images in the database using feature extraction methods. Similarity measurement techniques are used to find out and retrieve the similar images as shown in figure 2

·      Image Segmentation

In computer vision, image segmentation is the process of partitioning a digital image into multiple segments (sets of pixels, also known as super pixels). The goal of segmentation is to simplify and/or change the representation of an image into something that is more meaningful and easier to analyze. Image segmentation is typically used to locate objects and boundaries(lines, curves, etc.) in images. More precisely, image segmentation is the process of assigning a label to every pixel in an image such that pixels with the same label share certain characteristics.

The result of image segmentation is a set of segments that collectively cover the entire image, or a set ofcountours extracted from the image (seeedge detection). Each of the pixels in a region are similar with respect to some characteristic or computed property, such as color,intensity, or texture.

·      Multimedia Information Retrieval:

It is a research discipline of computer science that aims at extracting semantic information from multimedia data sources. Data sources include directly perceivable media such as audio, image and video, indirectly perceivable sources such as text, bio signals as well as not perceivable sources such as bio information, stock prices, etc. The methodology of MMIR can be organized in three groups:

1. Methods for the summarization of media content (feature extraction). The result of feature extraction is a description.

2. Methods for the filtering of media descriptions (for example, elimination of redundancy)

3. Methods for the categorization of media descriptions into classes.

Feature Extraction:

           In machine learning, pattern recognition and in image processing,feature extractionstarts from an initial set of measured data and builds derived values (features) intended to be informative and non-redundant, facilitating the subsequent learning and generalization steps, and in some cases leading to better human interpretations. Feature extraction is related to dimensionality reduction.

           When the input data to algorithm is too large to be processed and it is suspected to be redundant (e.g. the same measurement in both feet and meters, or the repetitiveness of images presented as pixels), then it can be transformed into a reduced set of features(also named afeatures vector). This process is calledfeature extraction. The extracted features are expected to contain the relevant information from the input data, so that the desired task can be performed by using this reduced representation instead of the complete initial data.

Input:-

           I=Image

Output:-

           Image Database = Similar Images from the search engine

Background Study:-

           The image retrieval retrieve’s the database images similar to the user’s query image(s), the retrieval system approximate the user’s similarity criteria, in order to identify the images which satisfy the user’s information need. User’s similarity criteria are represented by an image similarity model, which typically expresses the image similarity in terms of region similarities, and, in turn, each region similarity in terms of feature similarities. Elements of an image similarity model include:

1)     A collection of image features:

- e.g., color, texture, and shape - extracted from image Regions.

2)     The corresponding feature similarity measures, used to compute the individual feature similarities (of the image regions).

3)     Aggregation operator(s), used to aggregate individual feature similarities into region similarities, and region similarities into the overall image similarity.

4)     Weights, expressing the relative importance of individual features and regions. These elements are common for the existing image similarity models , almost without exception.

Description:

The concept of fast scanning algorithm is to scan from the upper-left corner to lower-right corner of the whole image and determine if we can merge the pixel into an existed clustering. The merged criterion is based on our assigned threshold. If the difference between the pixel value and the average pixel value of the adjacent cluster is smaller than the threshold, then this pixel can be merged into the cluster.

The concept of fast scanning algorithm is to scan from the upper-left corner to lower-right corner of the whole image and determine if we can merge the pixel into an existed clustering. The merged criterion is based on our assigned threshold. If the difference between the pixel value and the average pixel value of the adjacent cluster is smaller than the threshold, then this pixel can be merged into the cluster.

Indexing

1. Start

2. Select folder containing image dataset to be indexed.

3. for each image

4. Extract features EHD, HTD.

5. Store the extracted features in Database, with image path.

Searching

1. Select an image to be searched.

2. Upload image to the server.

3. Extract features of uploaded image (EHD, HTD).

4. Calculate Euclidean distance for the images stored in the database.

5. Rank result set according to the distances.

6. Display the Result of Images with Minimum Distances on Web Page to the User

7. Stop

Flow Chart

Activity Diagram:

Data Flow Diagram

Content Based Image Retrieval aims at avoiding the use of textual descriptions and instead retrieves images based on similarity in their contents for e.g. (Textures, Shape or Object).CBIR considers image globally and its contents. In CBIR searching is done based on contents of image rather than text.Content based image retrieval (CBIR) is an image retrieval technique for retrieving semantically-relevant images efficiently from an image database based on automatically-derived image features.