Abstract— signal may be modified by increasing sensitivity

Abstract— Video quality in general
plays a vital role in many fields like engineering, social areas as well as in
medical. Because of camera resolution and other factors many problems occurs
such as color enhancement, sharpening, pixel quality, visual effect and
especially zooming. Video super-resolution technique plays key role in high
resolution of display devices. This is class of technique that enhances the resolution of real time
video clips. Video super-resolution (SR) techniques are of
essential use for high-resolution display device due to the current shortage of
high-resolution videos. There are many algorithms available but, video SR still
remains a very challenging inverse problem under different conditions but, some
techniques are found useful for video quality like motion compensation and deep
residual learning etc. In this proposed system after zooming or stretching
video is clear based on super-resolution. They convert low resolution video to high resolution
of flexible video editing like de-interlacing, de-noising, de-blocking, color
correction, stabilizing, sharpening, visual effects and zooming etc.  In proposed work SR reconstruction applied specially for zooming
portions.

 

Keywords— Video super-resolution, Zooming,
frames etc.

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


order now

I.    
INTRODUCTION

The last several decades, there have been massive improvements
in modern digital cameras including resolutions and sensitivity. The quality of
videos is still limited. Firstly, videos have poor dynamic range. To capture
images of high dynamic range, most consumer cameras often rely on automatic
exposure control, but longer exposure time results motion blur. Secondly, the image
sequences taken with very low signal-to-noise ratio. The input signal may be
modified by increasing sensitivity of cameras. Various approaches are developed
for enhancing video proposed a noise-adaptive spatiotemporal filter that
considers both Poisson noise and false color noise of input videos. Because
their method aims only to videos slightly lower than normal lighting
conditions, enhancement of input dynamic range is omitted. The developed an
enhancement framework for low dynamic range video based on a virtual exposure
camera model. Their method includes the bilateral ASTA-filter (Adaptive Spatio-Temporal
Accumulation) and tone-mapping with a logarithmic function applied to a large
scale and detail features separately.

Because of camera
resolution and other factors many problems occurs such as color enhancement,
sharpening, pixel quality, visual effect and especially zooming. Super-
Resolution is a method to upscale video and images, i.e. increase resolution of
a video or picture (terms “upsize”, “up-convert” and
“uprez” are also widely used). To up-convert each frame, information
from this very frame and from
a number of neighbor frames is effectively used. If picture in video is not
changing too fast, then information from several frames is added to create a larger
and more detailed picture. There were many video SR algorithms
proposed, it is still
very difficult to handle all kinds of situations. SR techniques can be generally
classified into two categories single-image based and multi-frame based. Single-image based SR mainly interpolation based and example based methods.
Interpolation based methods are
of low computational cost but very limited restoration performance. High-resolution
videos add heavy storage and network
transmission burdens to the current video systems. In order to promote visual experience
sufficiently on high-resolution
display devices, video super-resolution (SR) is particularly essential. Video super-resolution is class of technique
that enhances the resolution of video system. Video SR techniques can also help video coding and decoding, face video
hallucination, video surveillance
systems, remote sensing systems,
intelligent robotic system, object recognition system, medical image analysis and
stereoscopic video processing.
Video
quality in general plays a vital role in many fields like engineering, social
areas as well as in medical Researchers and companies are now exploring efficient methods for accurate
video SR. The resolution is increase of video with motion based Super-Resolution method, where
each frame is upsized using information from a number of neighbor frames to
extract maximum details for outstanding results. The convert low resolution
video to high resolution of flexible video editing like de-interlacing,
de-noising, de-blocking, color correction, stabilizing, sharpening, visual
effects, zooming etc.

In proposed
work SR reconstruction applied specially for zooming portions. In this video
quality when video is zoom it will give accurate video zooming instead of
blurring pixel at a time of zooming video. Zooming is simply means
enlarging a picture in a sense that the details in the image became more
visible and clear. There are two different steps of zooming. The first step
includes zooming before taking an particular image. This is known as pre-processing
zoom. This zoom involves hardware and mechanical movement. The second step is
to zoom once an image has been captured. It is done with many different
algorithms in which we manipulate pixels to zoom in the required portion. An
image is zooming means changing the number of display pixels per image pixel
only in appearance. At zoom=1, there is one display pixel per image pixel. At
zoom=2, there are 2 display pixel per image pixel in both x and y this
enlargement is quantified by a calculated number or number is greater than one
called magnification when this number is less than one it refers to a reduction
in size called magnification. There are two types of zooming optical zoom and
digital zoom.                       

The
optical zoom is accomplished utilizing the development of the focal point of
camera. An optical zoom is really a concrete zoom. The after effect of the
optical zoom is obviously better than that of computerized zoom. In optical
zoom, a picture is amplified by the focal point such that the items in the
picture give off an impression of being nearer to the camera. In optical zoom
the focal point is physically reach out to zoom or amplify a object.
Computerized zoom is fundamentally picture handling inside a camera. A
computerized zoom, the focal point of the picture is amplified and the edges of
the photo got edit out. Because of amplified focus, it would appear that that
the object is nearer.  A computerized
zoom, the pixels got extend, because of which the nature of the picture is
traded off. A similar impact of computerized zoom can be seen after the picture
is taken through PC by utilizing a picture handling tool kit/programming, for
example, Photoshop. There are three strategies for zooming Pixel replication or
closest neighbor insertion, Zero request hold technique, Zooming K times.

 

 

                    II.
LITERATURE SURVEY

Image zooming is the process of
enlarging the spatial resolution of a given digital image. In this proposed
system a novel technique that intelligently modifies the classical pixel
replication method for zooming 1.This technique for zooming digital images.
This method decomposes a given image into layer of binary images, interpolates
them by magnifying the binary patterns preserving their geometric shape and
finally aggregates them all to obtain the zoomed image. Although the quality of
zoomed images is much higher than that of nearest neighbor and bilinear
interpolation and comparable with bi-cubic interpolation, the running time of Pixel Replication Technique is extremely fast like nearest neighbor interpolation and much faster
than bilinear and bi-cubic interpolation.

Video
 O/P

 

Frame to video conversion

 

Frame
Conversion
 

 

Video I/P

 

In video steganography Zero Order Hold (ZOH) technique. It is one
of the zooming technique, which is utilized for picture zooming. ZOH method is
applied for secret information embedded in the cover video. Steganography
used for secure transmission of secret message. Message transmitted over the
internet facing malicious attack. So, steganography is needed for secure transmission
of secret message2.
The new technique in digital image magnification with k times zooming method
which is called modified k time zooming method3. The magnification process is
created by using a method similar to k times zooming method by eliminating the
drawback that k times zooming method. The proposed work based on  modified k times image zooming method gives
the advantage of k times image zooming method have over nearest neighbor interpolation
and bilinear interpolation but by eliminating the difficulty of the k times
image zooming method of sorting at the end. Partial
Differential Equations (PDEs) 4 have become an important tool in image
processing and analysis. A PDE mode for image zooming is introduced in this
proposed work. This model exploits a higher order nonlinear partial
differential equation. The resulted nonlinear equation is solved by an explicit
finite difference schemes. Numerical results on real digital images are given
to show effectiveness and reliability of the proposed algorithm. These multiple-frames
super-resolution algorithms based on dictionary learning and motion estimation.
5. In proposed system, temporally recursive multi-frame SR algorithm, which
improve over single frame SR. A novel approach of algorithm is utilizing
consecutive video frames, rather than individual video frames. Which is
improves the performance of the video SR algorithms. Convolutional neural networks (CNN) 6 are a special type of deep neural
networks (DNN). They have so far been successfully applied to image
super-resolution (SR) as well as other image restoration tasks. In this
proposed system, consider the problem of video super-resolution. A  CNN is trained on both the spatial and the
temporal dimensions of videos to enhance their spatial resolution. Consecutive
frames are motion compensated and used as input to a CNN that provides
super-resolved video frames as output. While large image databases are
available to train deep neural networks, it is more challenging to create a
large video database of sufficient quality to train neural nets for video
restoration. A relatively small video database is sufficient for the training
of this model to achieve and even improve upon the current state-of-the-art.
Compare this proposed approach to current video as well as image SR algorithms.
There are many
algorithms, video SR still remains a very challenging inverse problem under
different conditions. There is deep residual learning for more accurate video
SR under different conditions including large and complex motions. The name of
method is motion compensation and residual net (MCResNet)7. In this method
optical flow algorithm is use for motion estimation and motion compensation as
a preprocessing step. A novel deep residual convolutional neural network (CNN)
to predict a high-resolution image using multiple motion compensated
observations. With deep residual learning, the representation capability of
CNNs is enhanced. The new residual CNN model preserves the low-frequency
contents and facilitates the restoration of high-frequency details. This method
is able to handle large and complex motion adaptively and readily applicable to
other video processing tasks.

 

                      III. PROPOSED
METHOD

 

 

 

Stretching /
 Zooming frames

 

 

 

 

 

             

 

Super-resolution    
algorithm (enhancement L to H)

 

 

 

 

 

   
           

                Fig.1 Block diagram  of  proposed
 method

 

Block diagram of video super-resolution
reconstruction as shown in Fig.1 In proposed work SR reconstruction applied specially
for zooming portions. This is class of technique that enhances the resolution
of real time video clips. They convert low resolution video to high resolution (HD) for
powerful and flexible video editing. Working principle of proposed system as
follows:

In this proposed
system input is video then covert into frames of same scenes. These images are
stretching or zooming. Super-Resolution algorithm applies for these
frames.  Then SR images combining together to form relevant information from
two or more images. Several images of same scene are combining together to form
high quality of image then convert into video. In frame to video conversion
these images into bunch of multiple frames and display the video.

 

Video I/P

A video
file format is a storing digital video data on a computer system. There are
different types of video file format AVI (Audio Video Interleave), FLV (Flash Video
Format),WMV (Windows Media Video), MOV (Apple QuickTime Movie), MP4 (Moving
Pictures Expert Group 4).

 

Frame conversion

 

Frame rate
may also be called the frame
frequency, and be expressed in hertz. The minimum frame rate to achieve   moving image is
about sixteen frames per second.

 

Super-Resolution
Algorithm

Super-resolution is technique which enhance of video or
image quality. In super-resolution technique covert low-resolution image to
high-resolution (LR to HR). HR means that pixel density within an image is high. Image magnification is one
of the basic image operations, and is widely used in many applications. Image magnification algorithms directly
affect the quality of image.
The image magnification is a conversion process from a low resolution image to
a high resolution image. The image magnification is essentially image
interpolation process. There have been a lot of practical image magnification
methods which have their own characteristics. The basic principle of image
magnification is to increase the image pixel number, so a low resolution image
is converted to a high resolution image. For
increasing resolution classical algorithms like nearest neighbor
interpolation, bilinear
interpolation and bi-cubic interpolation are used.

Frame to video conversion

In
this process again frames are combining together to form bunch of frames and
convert into video and display the video as output.

 

                                        IV.
RESULTS

There are three methods of zooming Pixel replication or Nearest neighbor interpolation, Zero
order hold method, Zooming K times. Zooming is simply means
enlarging a picture in a sense that the details in the image became more
visible and clear. There are two different steps of zooming. The first step
includes zooming before taking a particular image. This is known as
pre-processing zoom. This zoom involves hardware and mechanical movement. The
second step is to zoom once an image has been captured. It is done through many
different algorithms in which we manipulate pixels to zoom in the required
portion. Pixel replication is one of the zooming method and its result as
follows. Image is zoomed by replication factor-2. Image zooming with different file format jpg, tiff, bmp,  png  etc.

( a ) Original Image
        Tiger.jpg

 

Output image scaled by 2

 

  

Output image scaled by 2
 

 

( b)  Original Image
        Peppers.jpg
 

 

   

Output image scaled by 2
 

 

(c)  Original Image
        Parrots.tiff
 

 

   

Output image scaled by 2
 

 

(d) Original Images
        Flowers.png

 

   

 

 Fig.2 Image zooming with different file format
jpg, tiff, png etc

 

References 

1                
Kaeser Md. Sabrin and Md. Haider Ali, “An intelligent pixel
replication technique by binary decomposition for digital Image zooming,” in
dept. of computer science and engineering university of Dhaka, Bangladesh.

 

2                
Shashidhara H.N and Usha B.A, “Video steganography using Zero Order
Hold method for secured data transmission,” in Int. Journal of computer Applications, vol. 176, no. 5, 2017.

 

3                
Soumalya Banerjee and Anirban Saha, “Digital Image Magnification using
modified K-times Image zooming technique,” in Int. Journal of Innovative Research in comp &
com.eng.,vol.4,2016                         

 

4                
Ran Gao, Jin-ping song and Xue-cheng. Tai, “Image zooming Algorithm
Based on partial differential equations technique,” in International Journal of
Numerical Analysis and modeling, vol.6, no.2, pages 284-292, 2009.

 

5                
Qiqin  Dai, Armin Kappeler, “Sparse
Representation based multiple frame video super-resolution ,” IEEE trans on image processing, vol.11.no.
4. 2012.

 

6                
Armin Kappeler, Seunghwan, “Video super-resolution with convolutional
neural networks” IEEE trans. on
computational imaging,vol.2,no.2,2016.

 

7                
Dingyi Li, Zengfu
wang, “Video Super-Resolution via motion compensation and deep residual
learning,” IEEE trans. on
computational imaging2016.

x

Hi!
I'm Barry!

Would you like to get a custom essay? How about receiving a customized one?

Check it out