Image Comparison of WebP and JPEG, PNG in terms of Size,

Quality and PSNR.

MirTahmid Hossain(Taho1000)

Md. Mebah Uddin(mdud1000)

ABSTRACT: formats, we found that JPEG is its way back to webp

This paper describes an objective comparison of the considering the size and compression ratios of the
image quality of WebP and JPEG. Our approach is images keeping the image quality as metric.
basedon comparing the image quality, size and signal
noise ratio and plotting the results by estimating Introduction
thequality and size along with the signal noise
ratio.As we know, the Jpeg Format is a commonly Recent changesin communication and digital
used lossycompression technique for colour technology had led to find out several advancements
photographic images.It was popularly known in its content. However, digital images and videos are
worldwide and designed specifically for use with still demanding in terms of transmission bandwidth
highly detailed or realistic photos, it is applied to and storage space. There are so many image formats
rendered images and digitized photographs.WebP, available now. File extensions such as .jpeg, .bmp,
like JPEG, lets its users trade off image quality for file .gif, .png, and more can be seen after an image’s file
size. And like JPEG, WebP is also alossy format, so it name. Most of us will not consider it, thinking there is
doesn't perfectly reproduce an original image but tries nothing that important regarding these image formats.
to keep as true to the original as possible when These are all different and incompatible and these
viewed. This new image format also provides lossy image formats have their own pros and cons. They
compression for photographic images. In a large scale were created for specific, yet different, purposes. Web
study of several web images, WebP images were pages require JPG or GIF or PNG image types asall
39.8% smaller than that of jpeg images of similar the browsers can show these formats. On the web,
quality.The WebP format will be studied initially with JPG is considered as the best choice considering its
respect to encoding techniques used. Google claims smallest file size, with quality being less important
that WebP image codec has a better efficiency when than size, for picture images, and GIF is a common
compared to general JPEG formats. This will be format for graphical images.
checked by encoding a raw image in both JPEG and Lossy compression is necessarily needed to bring
WebP and comparing their SNR values. As the WebP these demands and deploy successfully on web etc.
does not have a viewer yet it cannot be verified In this report, we are planning to compare the most
subjectively. It can be intuitively changed into PNG popular file format JPEG with the recently
format andits SNR values can be compared with the introducedimage format WEBP by Google[1].This new
Jpeg. By comparing the JPEG andWebP image
image format also provides lossy compression for
photographic images. Google, having been released  Uses JPEG compression
webm for the sake of video, also shows its effect upon  Uses JPEG interchange format as a
still images. Overall, the goal is to introduce compressed image representation
the image compression technology VP8 in as many  Compatible with Windows or Mac or Unix
possibleways as we can. systems
 Standard colour space: one or three
JPEG components.
JPEG is an acronym for Joint Photographic Experts The JPG file is very small and can even
Group. It is termed as a standard method of compressedupto 1/10 of the size of the original data[2].
compressing photographic images. JPEG is the file The JPEG standard specifies the codec, which defines
format which employs this compression.Jpgs are how an image is compressed into a stream
termed as high colour images. In order to compress of bytes and decompressed back into an image, but
such kind of images, it is natural that some not the file format used to contain that
[2]
information has to be discarded . stream[2].The Exif and JFIF standards define the
The JPG format images are very well suited for the commonly used typical formats for interchanging of
internet. Bitmaps (.bmp), TIFF and other high colour JPEG-compressed pictures.
image formats can be compressed 10 to 20 times and
more when saved as Jpg format. Image files that WEBP
employ JPEG compression are generally called JPEG
files and are stored in variants of the JIF image WebP is a new image format introduced by Google. A
format. The most typical filename extensions for files WebP file consists of VP8 image data, and a container
employing JPEG compression are .jpg and .jpeg, which is based on RIFF[4]. The download repositories
though .jpe, .jfif and .jif are also used [3]. provided by Google source code consists of
Precompiled WebP utilities and library for
This format is mostly used for storing, viewing and
several operationg systems such as Linux,
transmitting photographs on the Internet.
Windows and Mac OS X. They include
Some facts are listed about JPEG format: the cwebp and dwebp programs to encode
and decode WebP files and libwebp that
 It is a well-establishedand popular standard could be used to add WebP encoding or
with numeroussoftware available, including
decoding into your programs.The
both proprietary and free software.
standalone libwebplibrary serves as a reference
 It isthe best format for photographs and
implementation for the WebP specification [4].
paintings of realistic scenes with smooth
Webmasters and web developers can use the WebP
variations of tone and colour.
image format to create smaller and better-looking
 JPEG compression artifacts are deployed and
images that can help make the web faster.
blended well into photographs with detailed
non-uniform textures, allowing higher The format was first announced in 2010 and is
compression ratios. supposed to be a new open standard for lossily-
compressed true-color graphics on the web, thereby
JPEG File Interchange Format features being presented as a direct competitor to the
older JPEG scheme, to which it is meant to compare provided. But WebP's promise might make this
favorably with the production of smaller files for worthwhile.
comparable image quality[1].
A. VP8:
WebP, like JPEG, lets its users trade off image quality
It is an open compression format which was
for file size. And like JPEG, WebP is also a lossy recently released by Google after it acquired ON2
format, so it doesn't perfectly reproduce an original technologies which was the previous owner of the
video codec. Google released the license for libvpx
image but tries to keep as true to the original as which is the reference format for VP8 under a BSD-
possible when viewed. This new image format also like license. Google along with forty other vendors
launched the WebM project to use VP8 as the video
provides lossy compression for photographic images. codec standard in HTML5. The VP8 codec is similar
In a large-scale study of several web images, WebP to H.264 with respect to many features like motion
estimation, predication, transform, etc. but there are
images were 39.8% smaller than that of jpeg images little differences which makes H.264 superior to VP8.
of similar quality[4].Google goals with WebP are Some of the features of VP8 are mentioned below [8].
a) Supports upto 3 reference frames
ambitious of replacing JPEG usage on web.
b) Variable block size motion compensation
c) Quarter pixel precision for motion
"When we took a bunch of images, recompressed
compensation
them from their current lossy formats into WebP, we d) Accurate DC transform
saw on average about 40 percent decrease in size, Secondary Hadammard transform performed
which is staggering," said Richard Rabbat[7] who is over DC transform for more compression
the lead product manager on Google's "make the Web
faster" effort. Shrinking images by that much is
COMPRESSION EFFICIENCY
particularly important considering that, by Google's
Compression efficiency is measured for lossless and
estimate, "65 percent of bytes on the Web are from
lossy compression. For lossless coding it is simply
images," he said[7].
measured by the achieved Scompression ratio for each
WebP is bitstream-compatible with VP8[5] and uses 14 one of the test images. For lossy coding the root mean
bits for width and height. The maximum file size for square error (RMSE) is used, as
[1]
WebP is 16383x16383 pixels . Webpcan be simply well as the corresponding peak signal to noise ratio[1]
termed as an AVI file with a single frame compressed (PSNR), defined as
with VP8. The compression of the VP8 is
significantly better than JPEG.
At a glance, WebPcan be termed as evolutionary step In terms of MATLAB, swe can deduce the formula of
to improve the balance between the quality and file Peak Signal Noise raito as follows
size especially for web images. As of now, this format 20*log10(255/sqrt(1/(M*N) *sum(sum((mainimage-
certainly has its challenges in terms of browser comparedimage). ^2))))
support and image viewer, but as more site owners
embrace the format and enjoy its associated speed
improvements, that is likely to change. IMPLEMENTATION
The experimental setup for our comparison needs
Supporting WebP now will require additional Matlab Software and a raw image along with the
development, since a fall-back JPEG should be encoders and decoders of webp format such as
cwebpencoder and dwebp decoder. Firstlywe have
chosen a raw image (lena.tiff) and converted into a Now we tried to find out the PSNR value of each
PNG image. Then by using cwebp encoder, we tried image from the above. After finding the values, we
to change it into webp format. Then after changing the plotted graphs considering the values of PSNR,
image into webp format, we then, changed it into Quality and Size of the image.
PNG using the dwebp decoder. Likely, we
createddifferent kinds of images from the decoded Results
PNG image from WebPconsidering the Quality After finding the values of PSNR of each image, we
constraint into 20 images with a difference of 5 in plotted graphs considering the values of Quality and
quality from 0 to 100. size opposite to the PSNR values.
Secondly, we have repeated the same process by The graphs that are obtained from the two image
choosingthe Raw image into JPEG first. Then we sub formats are listed below.
plotted the jpeg image into 20 different images Plots Obtained for WebP image format
considering the Quality metric of each image with a
difference of 5 from 0 to 100. So as of now we are
having 20 webp to png converted images and 20 jpeg
images with different image qualities.

X Axis-PSNR(db)
Y Axis-Size(kb)
160
140
120
100
80
size
60
40
20
0
30 35 40 45

Graph 1: PSNR vs Image SizeWebP

 X axis- PSNR(db)
Y axis-Quality
120

100

80

60
Quality
40

20

0
30 35 40 45

Graph 2: PSNR vs Quality WebP

X Axis - size (kb)

Y Axis - Quality

160
140
120
100
80
size
60
40
20
0
0 20 40 60 80 100 120

Graph 3:Sizevs Quality WebP

JPEG
160

140

120

100

80
size1
60

40

20

0
30 35 40 45

Graph 4: PSNR vs Image Size Jpeg

120

100

80

60
Quality1

40

20

0
30 35 40 45

Graph 5: PSNR vs Quality Jpeg

 X Axis - size (kb)
Y Axis - Quality

160
140
120
100
80
size
60
40
20
0
0 20 40 60 80 100 120

Graph 6: Image Size vs Quality Jpeg

above: high compression equals bad quality, low

We observed the image size of each file is almost compression equals large file size.
40% less than that of the Jpeg file. After finding the
PSNR values, we calculated the average values of
the two formats and the values are listed here. Conclusion
For WebP format, the average PSNR value WebP achieves overall higher compression than
obtained for 20 images is 33.65298 db. either JPEG or JPEG 2000. Gains in file size
For Jpeg format, the average PSNR value obtained minimization are particularly high for smaller
for 20 images is 36.94631 db. images which are the most common ones found on
the web.WebP is consistently better looking at low

When the images are heavily compressed the bitrates than JPEG, even when I visually inspect

quality drops, so you end up with graphs like the the results. At lower bitrates webp is resoundingly
better than JPEG.

References
[1] http://en.wikipedia.org/wiki/Webp
[2] http://en.wikipedia.org/wiki/JPEG
[3] W. Pennebaker, J. Mitchell, "JPEG Still Image Data
Compression Standard", Van Nostrand Reinhold,
New York, 1993. THE source to read about JPEG,
but also a nice presentation of basic material you
need to understand the rationale behind it
[4] http://code.google.com/speed/webp
[5] http://en.wikipedia.org/wiki/VP8
[6] http://en.wikipedia.org/wiki/Peak_signal-to-
noise_ratio
[7] http://news.cnet.com/8301-30685_3-20018146-
264.html
[8] WebM project, http://www.webmproject.org/