Posted on 72 Comments

Let AI Sharpen Your Photos

As photographers, we want to shoot sharp photos, period!  Our goal of getting sharp images is one of the major factors in the thousands of dollars we spend on good cameras and lenses with features like optical stabilization.  We’ve also invested time in learning techniques, tips, and tools for shooting sharp photos, like lugging around a sturdy but heavy tripod, setting proper shutter speed, learning about depths of field, proper focus, and so on and so forth.

Even with the best equipment and expert techniques, we sometimes still end up with images that could be sharper. That is why almost every photo editing software provides a “Sharpen” feature.  For example, Photoshop has no less than 6 different “sharpen” filters on my last count.

While we hope that getting a sharp picture will not be a problem when using a sharpening software, we know that’s not always the case.  Unless your image has a very small blur, existing software “Sharpen” functions do not work well.

Now let’s go into a technical explanation about how a given software sharpens your images. Up until now, the photo sharpening tools in software typically use one of two type methods: image filtering or deconvolution.

The first type of sharpening is Image Filtering. The basic idea is simple: blurry images lack spatial high-frequency components.  Therefore, applying a filter that boosts high-frequency will make the photo look sharper. Generally, this technique will works for photos with just a little blur.  Since this method is fast, most software sharpen tools use this method. In Photoshop, “Sharpen”, “Sharpen More”, “Sharpen Edge” and “Unsharp Mask” all use this type.  

Figure 1 (click to view details at 100%)

The Image Filtering method works well if you just need to sharpen the photo slightly, but if your photo has a bigger blur, it does not work well.  The problem is that when boosting high-frequency photo components, the tool will also also boost noise and artifacts in photos. Over using a sharpening filter that uses this method will create images with artifacts such as “fat edges”, “halo”, and noise amplification.

The second type of sharpening method is called Deconvolution.   Imagine you are looking through a camera at a star at night.  You will see, instead of a single bright point, a small blurry disk.  This disk represents the so-called point spread function. This function summarizes the blurring process as a mathematical operation called Convolution. Thus, removing blur is modeled mathematically as Deconvolution. This type of method was first developed primarily for astronomy and later for spy satellites. At its creation, normal photographers did not have access to this tech due to the computer speed and the stability of the algorithms.

The situation changed around 2010. There was a break-through in the deconvolution algorithm to make it feasible to use on a PC.  We were the first company to bring this tech to photographers when we released a Photoshop plugin, Topaz InFocus. This method works much better for small and moderate blurs.  It also works, to a certain degree, on blur due to camera shake or from a moving object. To this day, Topaz InFocus is still be considered to be one of the best sharpening tools on the market. Later, Photoshop also added “Smart Sharpen” and “Shake Reduction” based on deconvolution.

Figure 2. Large motion blur (click to view at 100%)

In addition to being much slower than the image filtering method, deconvolution needs to know the point spread function, which is very hard to estimate. It is also very sensitive to image noise and jpeg compression.  Therefore, it fails most of the time for large and complex blurs, such as one in Figure 2 (a). Since Hollywood’s crime dramas had (incorrectly) shown that they could usually turn very blurry photos into sharp ones to catch the bad guys, Topaz InFocus fell short of some of our customers expectations.  I took this quite personally and have been on the lookout for a silver-bullet ever since.

For eight years we did not find one…. not until recently. I felt we might finally have a shot to redeem ourselves thanks to the rapid development of Deep Learning in Artificial Intelligence (AI).

AI approaches the problem very differently. Instead of studying the mathematical model of the blur process and how to solve the inverse problem, we train an Artificial Neural Network with millions of blur-sharp image pairs.  The neural network will eventually “remember” what the sharp image should look like if it sees a blurry image. After months of training, the neural network could produce a sharper when image given an image it had not seen before. Figure 1 (c) and Figure 2 (c) are examples from our latest product, Topaz Sharpen AI.   Here is another example to sharpen an out-of-focus photo.

Figure 3. Out-of-focus blur (click for 100%)

Looking at Figures 1, 2, and 3 more carefully, you will find that Topaz Sharpen AI seems be able to create images with very fine details rather than simply sharpening the edges.  This is what makes Topaz Sharpen AI truly unique – it actually synthesizes convincing details even if the blurred image does not have any through the power of AI.

There is a reason that nobody has released an deep-learning sharpen AI tool for photographers so far.  It is quite an engineering challenge. For example, the neural network requires extremely high computation that makes regular PC pretty much unusable.  I am so proud that our team was able to overcome many challenges to bring Topaz Sharpen AI to you, from Dr. Acharjee’s  ingenious neural network architecture, to Image Processing Lead Bowen Wang’s efficient GPU neural engine development, to our whole product team’s application design.

You can try it yourself by downloading the 30-day free trial.  For anyone who owns Topaz InFocus or the complete bundle, it’s a free upgrade!

Does AI finally make blurry images a thing of the past? Not at all.  The example images I used are pretty much the best case scenarios to impress you with the AI.  You will find it is still a hit or miss situation sometimes. However, it is a huge step forward when comparing with any existing solution, including our Topaz InFocus. And best of all, we are just at the beginning of the Artificial Intelligence (AI) revolution.  You can be assured that there will be many more amazing things to come.

About the Author

Albert Yang founded Topaz Labs over 10 years ago, to form a company that adopts and implements the latest technology to introduce cutting-edge tools to the Photo market. With over 30 years of programming experience, he’s proud to offer his technical expertise to our users as the primary developer of our latest tools.

Posted on 29 Comments

The A.I. Evolution behind Topaz A.I. Clear

After a lot of effort from our team,  I am excited to share the latest version of A.I. Clear with you.  Topaz A.I. Clear is a software to remove photo noise and enhance sharpness. The update of A.I. Clear in Topaz Studio v1.12.5 is a completely new development. Not only does the new A.I. Clear do a better job of suppressing the noise, but further utilizes A.I. technology to transform noisy and blurry images into crisp and detailed images.

As a researcher in image processing for over 30 years, I cannot help but feel a little giddy about what our team has done.  Just take a look at the image above. Look at the clean removal of the noise. More importantly, look at the details in the output image!

It was a long journey to get here. Topaz Labs has always been at the forefront of image noise reduction. In 2010, Topaz Denoise had been already considered the best high-iso noise removal software in the market, a position it has maintained to this day.  The reason is that Topaz Denoise employed an algorithm that approached a theoretical upper-bound in terms of SNR (signal to noise ratio) increase. Therefore there were no much room for us, or anyone else for this matter, to improve it substantially. That is, until deep-learning in Artificial Intelligence (A.I.) emerged.

Hover over the blue plus signs to see more information.



Topaz Studio Ai Clear comes with three distinct neural networks, each trained specifically on a set of images with specific noise and detail requirements.


With one click, AI Clear automatically sets the noise reduction level using artificial intelligence to determine what setting is best for your image.


The Fastest of the three models, the Low setting is best for images with a range of subjects and a low to moderate amount of color noise. Cityscapes, landscapes, and architecture are great use cases for this model.


With Double the complexity of the Low model, the Medium setting is best for images with a defined subject and a moderate amount of noise and artifacts. This setting is particularly good on portraits and skin.


With double the complexity of the Low model, the High setting is best for images with a high amount of noise and artifacts. Evening shots, Long Exposures and Mid to High ISO images are cleaned and sharpened with this setting.


The new Enhance Sharpness feature allows you to manipulate the detail in your image to bring dull features back to life.


The Low setting is best for images with a smaller group of subjects. Works well on portraits, telephoto images, and astrophotography.


The High setting is best with a lot of small detail. This model works especially well with landscape photography.


If you find AI Clear to be removing a bit too much detail, increasing the value of this slider will bring back details that are removed by the adjustment.


Use this slider to increase or decrease the exposure of your result, higher values increase exposure, lower values decrease exposure.


Use this slider to increase or decrease the edge contrast of your image. Higher values increase edge visibility, lower values blend edges into surrounding detail .

A few years ago, the deep-learning based photo noise reduction in research started to produce a more promising result. A.I. Clear is the first commercial software that took advantage of this development.  It uses a “feedforward” neural network trained to suppress input image noise as much as possible. This type of method had proven itself to beat all known traditional methods in a research environment. Since Topaz Studio is an interactive app, the challenge was to discover a neural network architecture that could produce a good result at an interactive speed.  Through some pretty hardcore development work, we managed to deliver the first version of A.I. Clear, which produced better noise suppression than traditional methods without the need to manually fine tune every parameter. The reception from our customers was exceptional.

Then we began working on A.I. Gigapixel, the first deep learning desktop tool to enlarge photos. This long and detailed process vastly increased our understanding of A.I. based image enhancement. A.I. Gigapixel is based on a revolutionary concept in deep-learning called GAN (Generative Adversarial Network). It uses a “generative” neural network to synthesize plausible details that are non-existent in the input images. This process produces an unprecedented image enlargement result. The feedback for A.I. Gigapixel was so beyond our expectations and we knew we needed to use this technology for other purposes.

After the A.I. Gigapixel release, it was natural for us to try the same “generative” neural network idea on noise reduction and detail enhancement. So we did.  As expected, the “generative network” did not disappoint us.

But there was a catch. To produce the best result, this “generative” neural network needed a few orders of magnitude more “neurons and synapses” than the original one.  This much more complex network led to a very slow processing speed – an obvious no-go in an interactive app like A.I. Clear.  

Wanna Learn More?

Check out the AI Clear product page to learn more about AI Clear including pricing and to see more examples.

Eventually, our research scientist Dr. Acharjee found a pretty good compromise.  Through the work of our incredible development team, we managed to create a process that uses one-tenth the number of “neurons” than what we had started with. Not only did we match our previous speed – we exceeded it. Now, the speed is 2 to 3 times faster than the previous version of A.I. Clear, and we’ve improved image quality in most of our tests.  Still, for high noisy or blurry images it tends to synthesize unnatural-looking patterns (see image below) due to lack of enough “neurons” to remember the proper patterns.

While it was a pity that we could not put our best but admittedly slow neural network into A.I. Clear, we’re working on using this neural network in other ways. We’ve got a batch process image enhancer in the works that will use the no-compromise large neural network for those who want the very best possible result that A.I. technology can offer – regardless of how long it takes. A.I. Clear is impressive in what it can do, and even more capable technology is coming.

For now, I am excited to share the new A.I Clear with you. What do you think?

About the Author

About Albert Yang

Albert Yang founded Topaz Labs over 10 years ago, to form a company that adopts and implements the latest technology to introduce cutting-edge tools to the Photo market. With over 30 years of programming experience, he’s proud to offer his technical expertise to our users as the primary developer of our latest tools.

Posted on 19 Comments

A.I. Gigapixel – What does the “Reduce Noise and Blur” option do?


In the A.I. Gigapixel V1.1 update (released September 5, 2018), Our research scientist Dr. Acharjee has made substantial improvements on the deep learning models behind the scenes.   Those enhancements are reflected as replacing the “Enhancement” checkbox with a multi-level “Reduce Noise and Blur” options.

The difference is very subtle for most photos, for which the default setting (“Moderate”) will work well. For the highest quality result, you may want to try different “Reduction Noise and Blur” settings.

In this release, we have included 3 sets of neural networks tailored for images of different quality. The option selects one of the neural networks:


“None”:  This neural network was trained with clean images.  If your original photo is a well exposed RAW or is otherwise free of visible artifacts, this option will best preserve and create fine detail.  Here is an example of what the input image (top-left corner, 96×96 cropped from a large input image) should look and the enhanced result. Notice the input impage is pretty much free from high iso or other noise.

A.I. Gigapixel resize high quality RAW image to 400%

“Moderate”: This default option will suppress a moderate amount of image noise and JPEG compression artifacts and apply some sharpening. This setting is good for most consumer-level standard cameras and phone photos. In the example, you can see the noise in the input image, A.I. Gigapixel does a decent job resizing it to 400% while cleaning up the noise.

A.I. Gigapixel resize normal quality image to 400%

“Strong”: This option is for noisy, highly compressed, or otherwise artifacted images. This neural network applies the heaviest level of sharpening and noise reduction. Unfortunately, if too much information is missing or obscured by noise, A.I. cannot synthesize details properly (as in the example), and sometimes produce strange structures on details like faces. Alas, A.I. is not quite that magical when it comes to recreating faces yet. Therefore, do not use this setting on clean photos, that don’t have much noise.


A.I. Gigapixel resize noisy image to 400%

Since Gigapixel will not overwrite the original images, you can experiment with each setting to your heart’s content.

On a separate note, observant users may have noticed that the new update is slower when the image scale is less than 220%.  This is true. Dr. Acharjee has made substantial progress on the neural network archtecture for a considerable increase in image quality. According to feedback from the general community, speed takes a backseat to quality results.  For scales greater than 220%, different optimizations were implemented that will not result in longer processing times.

Happy A.I. Gigapixeling!

Feng (Albert) Yang – CEO, Topaz Labs

About the Author

About Albert Yang

Albert Yang founded Topaz Labs over 10 years ago, to form a company that adopts and implements the latest technology to introduce cutting-edge tools to the Photo market. With over 30 years of programming experience, he’s proud to offer his technical expertise to our users as the primary developer of our latest tools.