Image Compression Techniques Presented By: Palash RoyChowdhury Sai Krishna Madhavaram Mohammed Abdul Kareem
Image Compression TechniquesPresented By:Palash RoyChowdhury Sai Krishna MadhavaramMohammed Abdul Kareem
Introduction: Image Compression
> Data compression algorithms applied exclusively to images
> Reduces Cost for storage and transmission> Few compression types better suited for specific
image types > Adjustable parameters to improve compression
quality
Introduction: Image Compression
Introduction: Image Compression
> Direct cosine Transform (DCT) and Fast Frequency Transform (FFT)
> Lossy type Image compression algorithms > Joint photographic Experts Group (JPEG) is one of
the most widely used lossy type algorithms
Introduction: DCT
> The transformation of a two-dimensional matrix of pixel values into an equivalent matrix of spatial frequency components
> Once the equivalent matrix of spatial frequency components, known as coefficients, has been derived then any threshold can be dropped.
Introduction: DCT
> Our vision can percieve sharp edges better
> Such regions are considered “High frequency zones”
> DCT algorithms are useful for spearating High and Low fequency zones effectively
Direct cosine Transform (DCT)
> Pixels in the source image are divided into multiple blocks of 8*8.
> Each individual pixel position has a bit value between 0 to 255.
> The purpose of using DCT is essentially to replicate these 8*8 blocks.
Direct cosine Transform (DCT)
• Greyscale created by using cosine transforms.• Waves of different frequencies can be added and averaged
to form these respective shades.• This process also has centering of the pixel value• Cosine Transform is then applied using 8*8 bases of
Discrete Cosine Transform.
Direct cosine Transform (DCT)
Direct cosine Transform (DCT)
JPEG STANDARD QUANTIZATION MATRIX
Quantization
> Varies based on the Quantization matrix
> Allows us to fliter out coefficients based on frequency
> Large corporations like Adobe have their own proprietry matrix
Decompression
> DCT inverse gives back the shifted block but is centered at zero
> Adding 128 will recenter the pixel values
> Combining all the pixel blocks gives us back the compressed image.
Results - DCT
Fast fourier Transform
> FFT uses Decimation in Time algorithm to reduce the computational time for DFT (Direct fourier transform) computation.
> Details in the image are provided by high frequency components, but they are very susceptible to noise which causes spurious effects
> Many filters in image processing are based on FFT
Results - FFT
> For DCT, time taken to compress 5 images with different compression ratio:
Elapsed time is 3.124373 seconds> For FFT, time taken to
compress 3 images with different compression ratio:
Elapsed time is 7.295931 seconds
Thank you