6. different segments of video in the form of

6. Clustering & Indexing

this phase, the process of parsing the video and associating it with different
segments of video in the form of frames and it could be completely automatic or
manual. The indexing is fully automatic, and it is performed on both real-time
mode as well as stored videos. The indexing is done just simply contrasting the
current frame with previous generally known as key frame. The different frames
are only stored. Here also the comparison is based on the YUV based histogram
analysis. In indexing a threshold is decided for similarity of frames having a
difference of value less than threshold are considered to be same and only one
copy is saved and indexed. This difference is found through the cosine dissimilarity

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!

order now

has to be decided whether a shot boundary has happened, it is required to set a
threshold, or thresholds for the resemblance between adjoining frames. Cosine
similarity values above this threshold are register as real shot frontiers,
while values which are below this threshold are discarded. To accurately
segment video, it is necessary to balance the two apparently conflicting
points. The first one is to hamper identification of inaccurate shot boundaries
by setting a sufficiently high threshold level so as to sheathe the detector
from noise. The need to detect subtle shot transformations such as dissolves,
by making the detector sensitive enough to recognize gradual change.


7. Retrieval and Browsing

image processing, Content Based Image Retrieval is the process of retrieving
the desired query image from a large number of databases and combined them to
form metadata which are based on the contents of the image i.e. colour,
texture, shape and local features are some of the general techniques used for
retrieving a particular image from the images in the database 3. There are
few doubts which are related to retrieval first what is Image Retrieval? Firing
the Query as Image. Image Comparison i.e. Comparing Query Image with Referral Images
one by one. Return matched Images to user. Using key frames Image or video
frame itself as an input query Ex. Xcavator. Query by drawing query image using
provided drawing tools Ex. QBIC. System by keyword search by Image name or
description Ex. Google Image Search.

the retrieval of identical type of image or images, they first of all need to
be differentiated and if matched then their retrieval is done. The systems like
Fire, CIRES 5 perform the image retrieval based on only color histogram analysis.
Moreover there are also limitations for retrieval regarding the input query.
The standard search engines retrieves images based on only text query and not
as image itself as a query. On the other hand, in this system the image itself
can be given as a query and the similar kind of images are retrieved. The
retrieval is done on the basis of YUV histogram considering the intensity
factor and using the cosine similarity measure (Dcos value generation) formula
for comparison. In our system the partial as well as exact matched images are
retrieved based on user’s choice and accordingly the results are displayed with
comparison of Dcos value with standard threshold value. The image is given as
an input and the similar kind of images are retrieved from the database.
Comparison is done on the basis of content of image like color, intensity using
histogram analysis and cosine similarity measure. And based on threshold values
images similar to query image are retrieved successfully.