From
Send to

[KH Explains] Still confused about AI TVs? Look how machine learning improves picture quality

Samsung features 5-step 4K upscaling tech Q Engine, while LG offers 4-step color-focused Alpha 9

May 22, 2018 - 16:37 By Song Su-hyun
It’s been five months since artificial intelligence has officially expanded its territory to TVs.

But, few recognize the genuine role of AI in TVs, and are still confused about how AI TVs are different from existing smart TVs.

Consumers could have a better understanding of AI TVs by looking at major features adopted by the latest lineup of new TVs by South Korean TV makers Samsung Electronics and LG Electronics for 2018. 

AI on Samsung TV (Samsung Electronics)


Both Samsung and LG launched new TV products in March and April, and are highlighting that their products are the “real AI TVs.”

Of course, both companies equipped the TVs with their flagship voice-controlling AI platforms -- Bixby on Samsung’s QLED TVs and ThinQ on LG’s OLED TVs.

But the two archrivals are boasting their AI capabilities through picture quality more than other features, the biggest difference that an AI TV can have compared to previous smart TVs.

Samsung and LG both applied machine learning algorithms to improving picture quality.

Samsung has named the AI-enabled TV processor “Q Engine,” while LG is calling it “Alpha 9.”

Samsung’s Q Engine is a technology that upgrades standard-definition (400,000 pixels) and full high definition (2 million pixels) class image to 4K (8 million pixels) class.

It has applied a five-step algorithm to upscale any low quality images to 4K.
As the first step, the Q Engine analyzes image signals and then reduces noise in the image. Then, it examines details of the improved image after the noise reduction and upscale the picture quality of 4K. In the last stage, the image details are re-examined and further improved.

Samsung is also currently preparing to commercialize QLED TVs with an advanced upscaling technology to 8K.

LG’s Alpha 9 in OLED TVs has a four-step algorithm to analyze image signals and reduce noise. In the first two steps, the algorithms removes broken or noise parts of the image, and ease banding noise and blurring of colors in the third and fourth steps. 

LG Signature OLED TV (LG Electronics)


The Alpha 9 processor is particularly designed to represent accuracy of colors by seven folds compared to normal TVs.

“When you hear AI TVs, it is natural to think about controlling the TV with voice commands, but the real role of AI in those TVs is automatically improving picture quality,” said Park Joon-young, head of a machine learning startup.

“TVs with the machine learning algorithms, known as Machine Learning Super Resolution, learn mass database of images and can provide optimal colors and resolution out of the data,” he explained.

By Song Su-hyun (song@heraldcorp.com)