You are on page 1of 1

Implementation of skin detection using MATLAB and FPGA

Pankaj |14BEC0359| Prof. P. Jayakrishnan | SENSE


Motivation/ Introduction Results
Aim of this project was to implement skin detection using MATLAB and FPGA.
• Some color models are already present which give skin pixels. But these
MATLAB
models don’t give accurate skin pixels for variety of skins.
• Complex algorithms can be efficient with images. But when it comes to real
time processing, this will make system slow and performance will be
reduced. To overcome this bottleneck, skin color module should be efficient
and fast enough for real-time video processing

SCOPE of the Project Skin present in image


Original Image
Skin detection plays very crucial role in many applications such as image-based
steganography, face detection, pornography detection, human detection for
driverless cars, hand gesture recognition etc.

Every year, many loses their lives by suffering from skin diseases. If technical
science meets with medical science, then technology is a boon and all man-kind
will be blessed. To perform skin disease detection, skin detection is the first step.
Symptoms of disease can be identified in earlier stages. So, it will be a milestone if Indian Sub-Continent Skin Brazilian skin
we are able to detect cancer in earlier stage using this project.

Methodology
To get such a color model which will give more accurate skin pixels, there was
need to modify existing YCbCr color space. Modification of existing YCbCr space
and can be said as Y’Cb’Cr’ color space. RGB to Y’Cb’Cr’ color space
transformation is given as:
Australian skin European skin
𝑌′ 1 95 −37 112 𝑅 16 Detection Result Detection Result

𝐶𝑏′ = 128 −74 −100 𝐺 + 128


256
𝐶𝑟′ 45 80 −18 𝐵 128
No Yes
To get only skin pixels, we need
threshold limits, given below are what
we found after experiments,
90 ≤ 𝐶𝑏 ≤ 105,
143 ≤ 𝐶𝑟 ≤ 162 (Process flow for skin detection) #skin cancer found: 10 #skin cancer found: 2

Skin cancer: Seborrheic Keratosis


To get more precise location of
skin in image/real time, we FPGA
apply morphological operation
to remove unwanted skin pixels.
Erosion and hole filling will
reduce the background noise
and fill the missing pixels in the
detected regions found in
image.
(Process Flow used for MATLAB)
• Altera DE2-115 Cyclone IV EP4CE115F29C7
Real time using FPGA Skin detection with eye-glasses using FPGA
was used to implement this project on
FPGA with the use of Quartus 13.0.1. There
was requirement of VGA monitor and a
terasic camera.
• Terasic 5 MP camera- users can adjust the
exposure according to the light of the
surrounding area, Full resolution frame rate
up to 15 frame per second, 40 pins
connector.
(Implementation diagram) Real time skin detection using FPGA
• Each current frame from video
is captured by camera and is Conclusion/ Summary
sent to FPGA which processes
every frame and gives output Improved algorithm neither have huge number of computations nor have high
on VGA monitor via VGA cable. complexity. It is suitable for real-time implementation. There was some need of
• Spatial filtering is the similar modification in algorithm for software to implement it on FPGA.
process as erosion which can
be used in hardware. System proved its real-time implementation with no time lag and under various
(Process flow for implementation on hardware) • Sometimes sudden change in skin tones. Background colors, illumination and shadows are the main challenges
lightening will lead to flickering for performance of this system. But in normal conditions, efficiency is higher than
which can make VGA unstable.
previous works.
From our Verilog code, To reduce flickering, temporal
we need to make our new filtering can be used.
This can be used in applications which use skin detection as basic step. More
IP block. After following modification for algorithm in skin detection can be done.
everything step by step,
we will be able to
interface all components.
Contact Details
Then, Verilog code is pankaj2014@vit.ac.in
generated from Qsys.
That code is used for real- Acknowledgments/ References
time implementation on I am very grateful to my respected guide Prof. P. Jayakrishnan, Dean and HOD who
FPGA. (Process flow used in Qsys) helped me with their valuable suggestions throughout the project.

You might also like