Introduction: Most Detailed Explanations About Anti-aliasing, Gray Level and Image Blur Functions
The combination of anti-aliasing, gray level and image blur functionswill compensate for the pixel grain. However, most users are still confused about these functions. Here, follow us to get the most detailed explanations about anti-aliasing, gray level and image blur functions of CHITUBOX.
- Minimum feature size /pixel size of DLP/ MSLA(LCD) 3D printer
- Imaging principle of DLP/ MSLA(LCD) 3D printer
- How to compensate pixel grain with gray value
- Gray level and image blur function test
- Theory analysis about anti-aliasing, gray level and Image blur
Step 1: Minimum Feature Size /pixel Size of DLP/ MSLA(LCD) 3D Printer
DLP uses a digital projector to project individual images of each layer at once across the entire platform. Since the image of each layer is displayed digitally, it is composed of many square pixels.
MSLA(LCD) and DLP are almost the same but with the UV light coming from an array of LEDs shining through a LCD, not a projector. The screen acts as a mask, displaying only the pixels needed for the current layer.
As XY resolution is fixed for DLP/LCD systems, so there is a equation:
Width / X Axis = Height / Y Axis = Native Pixel Size
Taken ELEGOO SATURN as an example, it has a build area 192mm wide along the X axis. Divide 192mm (the X axis) by 3840 (the LCD width resolution in pixels), this will equal 0.05mm (50µm). Same thing for the Y-axis. Divide 120mm (the X axis) by 3400.That means the minimum XY-Axis feature size for this 3D printer is 50µm.
Driven by the design of the build platform, the Z resolution is different and need to be treated separately. The Z resolution (vertical resolution) is the minimal thickness of a layer which is actually a better indicator of surface finish. The smaller the layer height, the better the printed surface.
Layer height can be set in CHITUBOX in settings -- Print -- layer height. However, the minimum layer height is limited by the machine hardware. Generally, the minimum feature size of the Z-axis is specified in the specification. Commercially, the minimum layer height of a common DLP/ MSLA(LCD) is 0.05mm (50µm). In its specification, ELEGOO SATURN has a minimum Z-axis feature size of 0.00125mm (1.25µm).
So if calculated with the above minimum feature size, the smallest voxel that ELEGOO SATURN could produce was 50µm*50µm*1.25µm (X*Y*Z).
Step 2: Imaging Principle of DLP/ MSLA(LCD) 3D Printer (1)
Now that you know the minimum voxel that can be generated by DLP/ MSLA(LCD) 3D printer, let's explain how a DLP/ MSLA(LCD) 3D printer can reproduce image. First, we need to know that the LCD screen is divided into 50µm*50µm pixels in the XY direction.
The light source generates an image on the LCD screen. If the voxel is white, the light source at this location cures the liquid resin, forming a solid cube. If the voxel is black, then there is no light source at this location, and the resin does not be cured.
Step 3: Imaging Principle of DLP/ MSLA(LCD) 3D Printer (2)
Now, we have four 50µm*50µm*50µm cubes, and next to them, spaced at intervals, we have the same model made up of four 50µm*50µm*50µm cubes.
They can be 50µm, 100µ m, 200µm or more apart. But they have to be a multiple of 50µm apart. For example, we can't make them 175µm apart. Why? Because if our image is made up of pixels that are either pure white or pure black, there is no way to either expose them or not expose them to the fixed minimum grid on the LCD screen.
Step 4: Imaging Principle of DLP/ MSLA(LCD) 3D Printer (3)
So if it's gray between white and black, is it possible to do half exposure and half no exposure? So where is the half exposure? Left half, right half, front half, back half, top half, or bottom half?
The Ember team shared a study of how a single voxel is made. They lined the pixels up from dark gray (nearly pure black) to pure white, then used the corresponding light source to cure the resin.
The research showed that the gray level is lower than a certain value, and there would be no cured reaction. However, from dark gray to a certain critical brightness, semi-circular bubbles started to be formed in the previous layer. With the increase of brightness, the bubbles became larger and larger, and then the height started to increase slowly, and the lateral development was rapid. This suggested that the size of the voxel could be controlled by the brightness of each pixel.
Step 5: Imaging Principle of DLP/ MSLA(LCD) 3D Printer (4)
In practice, a gray pixel or hemivoxel would merge into adjacent voxels. So, if you have two white voxels from left to right plus one gray voxel (hemivoxel) plus one black voxel, you're going to have a hemivoxel near the two left voxels. By controlling the brightness of the pixel, any size of voxel can be produced.
Step 6: Imaging Principle of DLP/ MSLA(LCD) 3D Printer (5)
So can this research actually be applied to the actual 3D printing? Now let's look at this experiment. Five squares consisting of 10*10 minimum voxels (i.e. 500µm*500µm (0.5mm*0.5mm)) were equally spaced from left to right between the two dividing lines. From left to right, the top row of pixels in each square decreased from pure white to pure black, and the bottom row of pixels increased from pure black to pure white.
As can be seen from the printing results below, the missing distance of the top row of pixels of each square from left to right is about 0µm, 12.5µm, 25µm, 37.5µm and 50µm respectively due to the increase of the grayscale of the light source. It showed that it was completely possible to break the limit of minimum feature size by controlling the intensity of light source and achieve more accurate size control.
Step 7: How to Compensate Pixel Grain With Gray Value (1)
Suppose you want to print a T-shaped model (50µm*50µm*25µm (X*Y*Z) in the gray area shown above. Since both sides are composed of equably divided minimum voxels, in theory there will be no step pattern.
But if, adding a slope of 3.6° at the right of the inverted T as the yellow fields showing, then under the influence of black and white exposure, distinct step patterns appear as shown below.
Step 8: How to Compensate Pixel Grain With Gray Value (2)
By controlling the gray value of the slope so that each layer is 1.5µm narrower than the previous one, a relatively flat slope can be decreased.
If you change the angle to 75 degrees, the contrast will be even more pronounced. Because every 7-8 layers of 25µm, there is a 50µm step in the X direction. But if you compensate with grayscale, the slope will be relatively smooth.
Step 9: Gray Level and Image Blur Function Test (1)
First, let's take a look at the spheres' sides under the microscope, as shown below in the CHITUBOX preview.
In terms of the serious condition of surface water ripples, we get a ranking like this: non-antialiasing > anti-aliasing + gray level 8 > anti-aliasing + gray level 4 > anti-aliasing + gray level 0 (> denotes severity). Namely the surface quality of anti-aliasing + gray level 0 is the best.
Then, after image blur function is enabled, from the serious condition of surface water ripple, we get a ranking like this: anti-aliasing + gray level 4 > anti-aliasing + gray level 4+ image blur 2+ exposure time 2s. That is, the surface quality of anti-aliasing + gray level 4+ image blur 2 is better. At the same time, some horizontal stripes appear on the surface after the image is blurred. In order to determine whether it is influenced by the exposure time, we increase the exposure time of 2s to 3s, and then the horizontal stripes disappear. However, there is little difference between 2s and 3s exposure time as for surface water ripples.
Step 10: Gray Level and Image Blur Function Test (2)
Below, we compare the actual microscopic appearance of the T-shaped model with a slope, as shown below in the CHITUBOX preview.
In terms of the serious condition of the sawtooth, we get a ranking like this: non-antialiasing > anti-aliasing + gray level 8 > anti-aliasing + gray level 0 > anti-aliasing + gray level 4 (> denotes severity). That is anti-aliasing + gray level 4 has the best surface quality.
Then, after image blurring is enabled, the T-shape model with exposure time 2s also appears unknown horizontal stripes, and the horizontal stripes disappear after the exposure time is upgraded to 3s. In terms of the serious condition of the sawtooth, we get a ranking like this: anti-aliasing + gray level 4+ image blur 2+ exposure time 3s > anti-aliasing + gray level 4 + image blur 2 + exposure time 2s > anti-aliasing + gray level 4. That is the surface quality of anti- aliasing + gray level 4 is the best.
Step 11: Gray Level and Image Blur Function Test (3)
Finally, we compare the actual microscopic appearance of the sides of the triangular cone, as shown below in the CHITUBOX preview.
From the point of view of the seriousness of the sawtooth, non-antialiasing > anti-aliasing + gray level 8 > anti-aliasing + gray level 4 > anti-aliasing + gray level 0. Namely anti-aliasing + grayscale level 0 has the best surface quality.
Then, after the image blur is enabled, unknown horizontal stripes also appear in the triangular cone of 2s exposure time. After the exposure time is upgraded to 3s, the horizontal stripes disappear. In terms of the seriousness of the sawtooth, anti-aliasing + gray level 4 > anti-aliasing + gray level 4+ image blur 2+ exposure time 2s > anti-aliasing + gray level 4+ image blur 2+ exposure time 3s. That is, anti-aliasing + gray level 4+ image blur 2+ exposure time 3s has the best surface quality.
The results of this test apply to the maroon rigid resin used in the test. In addition, the post-processing may also affect the actual surface conditions. In practical application, it needs to be set according to the actual situation of the machine, especially the light source, as well as the characteristics of the resin material. Anti-aliasing, gray level and image blur are used in combination and collocation. The optimal parameters will be different under different actual conditions.
Step 12: Theory Analysis About Anti-aliasing, Gray Level and Image Blur (1)
After comparing the tests, now let’s further analyze the sliced file. For the convenience of illustration, we take a cylinder of 0.3*0.3*0.3mm, which is divided into 6 parts of 50µm (0.05mm) according to the diameter calculation. It we set the layer thickness to 0.025mm, then the total number of layers is 12.
Although the images of the 12 layers should be the same circle, in the actual printing, images of each layer actually change slightly as the DLP/ MSLA(LCD) can only generate squared voxels. In order to control variables, we take the images of the sixth layer for comparison.
Step 13: Theory Analysis About Anti-aliasing, Gray Level and Image Blur (2)
First of all, let’s see the difference among Non-antialiasing, level 2 anti-aliasing, level 4 anti-aliasing, and level 8 anti-aliasing (Turning on anti-aliasing will synchronously turn on the gray level. Here the gray level is uniformly set to 0). From the perspective of pixel distribution, the maximum number of non-antialiasing pixels on the Y-axis is 6, while the other anti-aliasing pixels are all 8. In terms of pixel density, there are only 28 pixels for non-antialiasing, while all the other anti-aliasing have 36 pixels. In terms of gray value (not counting black and white), there is no gray value for non-antialiasing. There is one gray value (127) for level 2 anti-aliasing, two gray values (127, 191) for level 4 anti-aliasing, and two gray values (127, 223) for level 8 anti-aliasing.
Step 14: Theory Analysis About Anti-aliasing, Gray Level and Image Blur (3)
Here's an added bonus:
The gray level is the gray value, which is the concept of brightness, 0~ black, 255~ white, depending on the color depth range of 0~255. (Must be between 0 and 255). CHITUBOX classifies this color range into nine levels from 0 to 8, with 0 bits black, 8 white and 1 to 7 gray in the middle.
Then compare the difference of anti-aliasing + gray level 0, anti-aliasing + gray level 4, and anti-aliasing + gray level 8. From the perspective of pixel distribution, the longest pixel distribution on the Y-axis of all the three has 8 pixel points. From the perspective of pixel density, gray level 0 has 36 pixels, while gray level 4/8 has 40 pixels. In terms of gray value (not counting black and white), there are six gray values (124, 126, 128, 220, 222, 224) for gray level 0, six gray values (126, 128, 188, 190, 236, 238) for gray level 4, and no gray value for gray level 8.
In addition, pay attention to the gray level problem. When the gray level goes 8, all pixels are pure white with a gray value of 255 (here the algorithm's gray value is 254). It can be seen that, after turning on the gray level, not only the Y-axis pixel distribution, pixel density and gray value are affected, but also the pixel brightness. When choosing gray level, be careful when choosing higher levels.
Step 15: Theory Analysis About Anti-aliasing, Gray Level and Image Blur (4)
Image blur is similar to the feathering function in Photoshop (make the inner and outer connection of the selection become blurred, playing a gradual role to achieve the natural cohesion).To put it simply, gently wiping pencil lines with a hand or eraser will cause the lines to blur and render. Image blur is a similar operation. The higher the level, the more blurry it is.
Finally, let’s compare the difference between anti-aliasing + gray level 4+ image blur level 2 (simplified as "image blur level 2" below) and anti-aliasing + gray level 4+ image blur level 4 (simplified as "image blur level 4" below). From the perspective of pixel distribution, the longest pixel distribution of image blur level 2 on the Y-axis has 9 pixels, while image blur level 4 has up to 11 pixels.In terms of pixel density, image blur level 2 has 54 pixels, while image blur level 4 has 93 pixels. Look from the grey value (not counting black and white), image blur level 2 has 11 grey values (126, 142, 170, 186, 190, 198, 202, 234, 238, 246, 250), while there are 20 for image blur level 4 (130, 138, 140, 142, 148, 150, 152, 156, 158, 160, 162, 176, 180, 184, 188, 206, 208, 216, 234, 244).
It should be noted that when the image has a higher level of blur, the grayscale starts to render inward at the same time, resulting in the original white pixels in the central area becoming gray, which is likely to cause insufficient central light source and lead to the problem of curing. When selecting image blur level, be careful when selecting higher levels.
In the CHITUBOX algorithm, the sequence of priority for calculation is image blur - gray level - anti-aliasing. Image blur controls the degree of feathering (that is, the number of edge transition pixels), gray level controls pixel brightness, and anti-aliasing controls the number of transition gray values. The combination of the three will compensate for the pixel grain.