Paras Maharjan profile picture

Hello, I'm

Paras Maharjan

Computer Vision Researcher

My LinkedIn profile My Github profile My Google Scholar profile

Uncover my

Academic Journey

Experience icon

University of Missouri - Kansas City

Aug 2017 – Dec 2019

Master's in Electrical and Electronics Engineering

GPA: 4.0/4.0

Experience icon

University of Missouri - Kansas City

Jan 2020 – July 2025

Ph.D in Electrical and Computer Engineering

Supervisor: Dr. Zhu Li

GPA: 4.0/4.0

Explore My

Experience

Experience icon

Sony Corporation of America

Sept 2025 – Present

AI/ML Computer Vision Research Intern

Working on solving computer vision problem using AI/ML.

Experience icon

Multimedia Computation and Communication Lab (MCC)

June 2018 – July 2025

Graduate Research Assistant

Worked on low-light image denoising, image/video compression, Fast SIFT keypoint detection, image restoration like deblurring, deblocking, etc.

August 2021 – Dec 2023

Instructor

Engineering Computation (ECE 216/217)


Teaching Assistant

Multimedia Communication(ENGR 5578), Remote Sensing (AS 280), Computer Vision (ENGR 5582)

Experience icon

Atombeam Technologies

Feb 2024 – July 2025

Intern Research Scientist

Architected an end-to-end image compression solution utilizing deep learning techniques for complex-valued SAR images, achieving a 28% reduction in inference time.


Optimized channel context modeling through feature grouping based on latent energy, leading to a 26.5% BD-rate improvement over VVC/H.266-based SAR image compression.

Experience icon

Dolby Laboratories

May 2022 – Aug 2022

Video Processing Research Intern

Formulated a local gamma correction algorithm for adaptive face brightness adjustment in video conferencing scenarios.


Developed a machine learning-driven technique to determine the optimal tuning parameter for background exposure correction.

June 2020 – Sept 2020

Image Processing Engineering Intern

Accelerated real-time sports recognition and image segmentation using deep neural networks, achieving a 3× speedup for sports recognition and a 30% improvement in segmentation.

Experience icon

Kuaishou Technology

Jun 2021 - Aug 2021

Research Intern of Image/Video Algorithms

Constructed a deep neural network for image deblocking and deblurring, leveraging DCT decomposition for enhanced transform-domain processing.

Sept 2020 - Jan 2021

Research Intern of Image/Video Algorithms

Implemented deep learning-based image denoising techniques for real-world noisy images.


Proposed a dual-input-dual-output network with a dual loss function that processes noisy raw and ISP-processed noisy sRGB images to generate clean sRGB outputs.

Experience icon

Poly Inc.

May 2022 – Aug 2022

Video Processing Research Intern

Prototyped multi-camera image fusion techniques to improve image quality in real-time applications.


Developed an algorithm for multi-camera systems incorporating geometric calibration, image registration (SIFT), parallax correction, and alpha fusion-based blending for seamless video transitions.

Browse My Recent

Publications

Project 1

Improving Extreme Low-Light Image Denoising via Residual Learning

Project 2

DCTResNet: Transform Domain Image Deblocking for Motion Blur Images

Project 3

Fast LoG SIFT Keypoint Detector

Project 1

Complex-valued SAR Image Compression: A Novel Approach for Amplitude and Phase Recovery

Project 2

End-to-End Compression of Complex-Valued SAR Images

Project 3

Distributed Polarimetric SAR Compression with Joint Deblocking Using Side Information

Arrow icon

Learn About My

Patents

Systems and methods for neural network based data compression

US Patent: 12224777

Real time discrete cosine transform image and video processing with convolutional neural network architecture

US Patent: 12198304

Adaptive face brightness adjustment for images and video

Worldwide Patent: WO2024112375A1

System and method for learning-based lossless data compression

US Patent: 12119848

Methods and devices for joint sensor and pixel domain denoising

US Patent: 11967047

Controllable lossy compression system using joint learning

US Patent: 12199643

Upsampling of compressed financial time-series data using a jointly trained Vector Quantized Variational Autoencoder neural network

US Patent: 12229679

System and methods for upsampling of decompressed genomic data after lossy compression using a neural network

US Patent: 12224044

System and methods for upsampling of decompressed time-series data using a neural network

US Patent: 12068761B1

Get in Touch

Contact Me