3D-CNN Autoencoder for Plenoptic Image Compression

Zhong, Tingting; Jin, Xin; Tong, Kedeng

doi:10.1109/vcip49819.2020.9301793

Cited by 7 publications

(7 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The CNN-based angular super-resolution model of [22], which uses a set of hand-crafted features for learning the disparity, forms a part of most of these approaches. While all the above learning-based approaches have a hybrid structure with dependence on multiple components, [23] proposes a simple 3DCNN autoencoder. The 3D convolution enables learning in the pseudo-temporal dimension and hence the model has an input data structure awareness.…”

Section: Prior Workmentioning

confidence: 99%

“…Table II shows RGB PSNR values averaged across the 4 lenslet images available in the JPEG Pleno Light Field dataset [35]. Comparing with the only other end-to-end learning-based method [23] in terms of the reported RGB PSNR values there seems to be no advantage. However, there is a gain in terms of the reported encoding speed.…”

Section: Evaluation 1) Metricmentioning

confidence: 99%

“…It is important to note that we trained our model for much lower than 1M steps as compared to 5M steps in [25] with similar models for 2D images. At the same time, our training dataset is much smaller as compared to [23] which uses more than 6000 light field images and a patch size of 256 × 256 for training. The loss used is simple MSE as compared to a perceptual loss in [21].…”

Section: Evaluation 1) Metricmentioning

confidence: 99%

See 2 more Smart Citations

Learning-Based Practical Light Field Image Compression Using A Disparity-Aware Model

Singh¹,

Rameshan²

2021

Preprint

View full text Add to dashboard Cite

Light field technology has increasingly attracted the attention of the research community with its many possible applications. The lenslet array in commercial plenoptic cameras helps capture both the spatial and angular information of light rays in a single exposure. While the resulting high dimensionality of light field data enables its superior capabilities, it also impedes its extensive adoption. Hence, there is a compelling need for efficient compression of light field images. Existing solutions are commonly composed of several separate modules, some of which may not have been designed for the specific structure and quality of light field data. This increases the complexity of the codec and results in impractical decoding runtimes. We propose a new learning-based, disparity-aided model for compression of 4D light field images capable of parallel decoding. The model is endto-end trainable, eliminating the need for hand-tuning separate modules and allowing joint learning of rate and distortion. The disparity-aided approach ensures the structural integrity of the reconstructed light fields. Comparisons with the state of the art show encouraging performance in terms of PSNR and MS-SSIM metrics. Also, there is a notable gain in the encoding and decoding runtimes. Source code is available at https://moha23.github.io/LF-DAAE.

show abstract

Section: Prior Workmentioning

confidence: 99%

Section: Evaluation 1) Metricmentioning

confidence: 99%

Section: Evaluation 1) Metricmentioning

confidence: 99%

See 1 more Smart Citation

Learning-Based Practical Light Field Image Compression Using A Disparity-Aware Model

Singh¹,

Rameshan²

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Inspired by this, researchers have explored many deep learningbased video coding methods [15][16][17]. For light field image coding, some end-to-end coding frameworks have also been proposed, and preliminary results have been achieved [18][19][20][21]. Specifically, the first step involves the extraction of light field image features.…”

Section: Introductionmentioning

confidence: 99%

End-to-End Light Field Image Compression with Multi-Domain Feature Learning

Ye,

Li,

et al. 2024

Applied Sciences

View full text Add to dashboard Cite

Recently, end-to-end light field image compression methods have been explored to improve compression efficiency. However, these methods have difficulty in efficiently utilizing multi-domain features and their correlation, resulting in limited improvement in compression performance. To address this problem, a novel multi-domain feature learning-based light field image compression network (MFLFIC-Net) is proposed to improve compression efficiency. Specifically, an EPI-based angle completion module (E-ACM) is developed to obtain a complete angle feature by fully exploring the angle information with a large disparity contained in the epipolar plane image (EPI) domain. Furthermore, in order to effectively reduce redundant information in the light field image, a spatial-angle joint transform module (SAJTM) is proposed to reduce redundancy by modeling the intrinsic correlation between spatial and angle features. Experimental results demonstrate that MFLFIC-Net achieves superior performance on MS-SSIM and PSNR metrics compared to public state-of-the-art methods.

show abstract

“…Select 80% of each class in the dataset at random as the training sample and the remainder as the test data Images and providing accuracy of 99.94%, 99.96%, and 99.99% respectively. Tingting Zhong Et.al [5] present an Adaptive Data Structure i.e a 3D convolutional (DSA-3D) auto encoder that adapts to the bone frame of pictures from a light field camera. The experimental assessment is performed using the EPFL Light Fields dataset, which comprises 118 plenoptic pictures, as well as the dataset created by our laboratory, which has 6000 plenoptic images captured by Lytro Illum.…”

Section: Introductionmentioning

confidence: 99%

Oral Cancer Detection in its Nascent Stages Using 3D CNN: A Survey

Sagari¹,

Solanki²,

Goel³

et al. 2023

IJSREM

View full text Add to dashboard Cite

Abstract—oral cavity cancer also called squamous cell carcinoma is the sixth most frequent cancer overall, it is a rising public health concern that is spreading across the globe. If not detected in earlier stages, it might be severe generally, and underlying lesions induce it. Therefore, it might be difficult to locate the lesion area and if not cured in its nascent stages, it might reach stage IV cancer, called the metastatic stage. To come up with a solution, we tend to use the three- dimensional CNN processes malignancies with image data and 3DCNN structures to classify benign and aggressive oral tumors. By identifying the kind of tumor region susing various kinds of knowledge here on periodic alterations in tumor enhancers, CNN was utilized in Test with visual categorization improvement rates. Analyzed, categorized, and future study areas are identified. The datasets utilized in the review articles are covered in detail, as are the pre- processing methods, the deep learning frameworks employed, etc. It presents a comparison of the same. Index Terms—Oral Cancer, Oral lesions, Oral Cavity, 3D CNN, Pre-processing, tumors, malignant, deep-learning

show abstract

3D-CNN Autoencoder for Plenoptic Image Compression

Cited by 7 publications

References 10 publications

Learning-Based Practical Light Field Image Compression Using A Disparity-Aware Model

Learning-Based Practical Light Field Image Compression Using A Disparity-Aware Model

End-to-End Light Field Image Compression with Multi-Domain Feature Learning

Oral Cancer Detection in its Nascent Stages Using 3D CNN: A Survey

Contact Info

Product

Resources

About