fMRI Exploration of Visual Quality Assessment

TL;DR

This work probes the neural basis of visual quality perception using fMRI by contrasting image quality assessment (QA) with content classification (CC) across tasks and image qualities. It combines univariate activation analyses and task-based functional connectivity to reveal that QA engages both detailed visual processing and higher-order cognitive networks, with QA and CC recruiting distinct patterns of brain activity and connectivity. The study also demonstrates that the brain employs different strategies when processing high- versus low-quality images, relying on early visual cortices for high-quality content and mobilizing broader visual and cognitive circuits for degraded inputs. These findings provide neuroscientific insight into image quality perception and offer a data-driven foundation for developing objective IQA algorithms that align with human neural processing.

Abstract

Despite significant strides in visual quality assessment, the neural mechanisms underlying visual quality perception remain insufficiently explored. This study employed fMRI to examine brain activity during image quality assessment and identify differences in human processing of images with varying quality. Fourteen healthy participants underwent tasks assessing both image quality and content classification while undergoing functional MRI scans. The collected behavioral data was statistically analyzed, and univariate and functional connectivity analyses were conducted on the imaging data. The findings revealed that quality assessment is a more complex task than content classification, involving enhanced activation in high-level cognitive brain regions for fine-grained visual analysis. Moreover, the research showed the brain's adaptability to different visual inputs, adopting different strategies depending on the input's quality. In response to high-quality images, the brain primarily uses specialized visual areas for precise analysis, whereas with low-quality images, it recruits additional resources including higher-order visual cortices and related cognitive and attentional networks to decode and recognize complex, ambiguous signals effectively. This study pioneers the intersection of neuroscience and image quality research, providing empirical evidence through fMRI linking image quality to neural processing. It contributes novel insights into the human visual system's response to diverse image qualities, thereby paving the way for advancements in objective image quality assessment algorithms.

Figures (5)

• Figure 1: Illustration of two main procedures of the proposed pruning method: train the decorators and get the mask for pruning in the split state, and retrain the model with the mask to recover its performance with merged convolutional layers.
• Figure 2: Illustration of the distribution of average response times for images of different quality levels and content on the quality assessment and content classification task.
• Figure 3: The effects of task and quality condition on blood-oxygen-level-dependent (BOLD) activity. (a) brain regions showing increased and decreased activities during the QA task compared with the CC task. (b) Regions of the brain show increased and decreased activity when viewing high-quality images compared to low-quality images.
• Figure 4: Brain regions demonstrating significant activation across both "High-quality vs Neutral-quality" and "Neutral-quality vs Low-quality" contrasts scenarios. (a) and (b) respectively illustrate regions of positive and negative activations under the two distinct contrasts. The hue of yellow has been employed to highlight brain territories that exhibit activation solely within one contrast, whereas the color magenta represents regions showing consistent activation across both contrasts. The criterion for statistical significance, denoting activation, has been conservatively established at a threshold level of $p \textless 0.001$, FDR corrected at voxel level ($p \textless 0.05$). (c) - (f) respectively illustrate the distribution of Blood Oxygen Level Dependent (BOLD) signal intensity within subregions of the Lingual Gyrus, Middle Occipital Gyrus, Superior Occipital Gyrus, and Middle Frontal Gyrus across different image quality levels for each of the 14 participants. The vertical axis in these plots has been normalized to allow comparison, with the specific demarcation of these subregions derived from the common activation regions identified in (a) and (b).
• Figure 5: Functional connectivity analysis results on quality assessment and content classification tasks. Shows the comparison of functional connectivity patterns in the task state and the resting state. Detailed brain network connectivity is shown on the left, the color bar on the left network diagram represents the t-value of the connection. Positive connectivity (positive correlation) is represented by the red line and negative connectivity (negative correlation) is represented by the blue line in the figure on the right.