How to increase clarity of a photo when the hill is hazy

A storm door and a hazy hill are veils, not blur from a handshake. Denoise the photo and keep the ridge line where it was.

The picnic frame is from July 2016, a borrowed camera, and the message that came with it was how to increase clarity of a photo. Everyone at the table is sharp. The plastic cups, the fence wire, the grain of the table are all there. A mile back, the hill is a pale shape and the sky is almost the same pale. My mother remembers a path on that hill. In the JPEG the path is a rumor. She wanted the path, not a new face and not a postcard sky.

The file is 5472 by 3648 pixels, about 8 MB, JPEG. EXIF says f/8, one four-hundredth of a second, ISO 200, a normal lens around 35mm. Those settings are not a blur recipe. One four-hundredth will stop a cup being set down. f/8 at that focal length, focused on the fence, leaves a distant hill softer only by optics and air, and the softness I see on the ridge is not the disk of missed focus. Edges do not bloom. Contrast collapses. The hill and the sky sit about 12 levels apart on a 0 to 255 scale in a sample I took along the ridge line. The foreground grass, in a patch the same size, swings about 40 levels from blade to blade. The information is local contrast, and the distance has less of it.

I tell her that before I touch a slider. How to increase clarity of a photo, for this afternoon, means giving the hill some of its midtone separation back without sanding the people and without inventing a mountain range. The path is the test. If a switchback appears as a faint line and the sky stays quiet, I stop. If the sky fills with grit, I went too far, even if the path looks brave.

TL;DR. The picnic is sharp and the hill is hazy. EXIF is f/8 at one four-hundredth and ISO 200, and the fence wire is about 1 pixel clean, so this is not a focus job. The ridge and the sky differ by about 12 levels. A masked local-contrast pass, plus a dehaze of about 15 on the distance only, brought that gap to about 22 and revealed a path about 2 pixels wide. Faces stayed masked. A global clarity pass mottled the sky and haloed hair. I did not sharpen the ridge. The weather is still weather.

Focus, checked on purpose, so I do not deblur a mile of air

I zoom to the fence before I believe the EXIF. The wire is a dark line against dry grass, and the transition is about 1 pixel. A missed focus would put a gray ramp several pixels wide on that wire. It does not. The near edge of the table is the same. A nephew's eyeglass frame is crisp enough to show a small bend in the temple that I can confirm on his face in the room with me. The camera focused. The shutter did its job. ISO 200 is not a noise crisis.

The ridge, zoomed the same way, does not show a ramp around a sharp edge that failed. It shows no edge. The hill melts into the sky over a band maybe 30 pixels tall, and inside the hill the tones are flat. That melt is what haze does. Scattering adds light to the shadows and lifts the darks. A deblur model looking for an edge to restore will look at that band and may draw a ridge line sharper than the air allowed, a dark contour that was not a contour. I tried a small deblur on a crop of the hill, on a copy. It drew that contour and added a suggestion of trees I cannot find in the original even as a texture. I deleted the copy. There is a fence to prove the lens was fine. I do not need a fictional tree line to prove it again.

The path is the one structure my mother swears was visible to her eye. Eyes compress haze differently than a JPEG does. I believe she saw a path and I also believe the file barely has it. After the masked pass it is a light line, about 2 pixels wide in the 5472-wide file, switching back once. I can follow it for about 400 pixels before it fades. I did not extend it past where the samples support it. A path that continues into the next county is a drawing.

Local contrast, masked, with numbers I re-sample

Clarity, in the tool I use, is local contrast. It darkens local darks and lightens local lights inside a radius, without shifting the whole picture brighter. That is different from a contrast slider on the full frame, which would push the already-bright sky to the wall and crush the cups. I set the radius large enough to see the hill's folds, on the order of the folds themselves, and I kept the amount modest. Then I sampled again. Ridge versus sky moved from about 12 levels to about 22. The path became visible at 50 percent zoom and remained a line, not a road, at 100 percent.

The mask is the work. People at the table are painted out. The sky is painted out. The hill and the far trees are painted in, with a soft edge along the fence so the fence does not get a processed side and a raw side. The first test, which I ran with no mask because I was curious, did three bad things. Hair got a light rim about 1 pixel wide. Freckles on the nephew jumped in contrast until they looked like specks of dirt. The sky, which had been smooth at ISO 200, broke into faint blotches, especially in the corner away from the sun. I kept that test under a name that means failure. The delivered file is the masked one.

Dehaze is the partner, not the same knob. It tries to remove a veil, and it also saturates and darkens. On the hill alone, an amount I would call 15 on a 0 to 100 scale put density back in the far trees without a new color. At 40, the trees went cyan at the edges and the dry grass went an angry yellow-green. Skin, in the failed full-frame test, went dull and slightly magenta in the shadows. Amount 15, distance only, plus the local contrast, is the combination that moved the sample from 12 to 22 without a new climate. I wrote 15 in the notes so I do not "improve" it next week from memory.

The path is the stop, not the sky

When the switchback showed up, I stopped adding amount. Further clarity made the 2 pixel line darker and did not reveal a rock or a person on it. There is no person on it. I checked by raising contrast in a temporary view, which is a way to hunt for hidden shapes, and I saw noise, not a hiker. Temporary views get thrown away. They are flashlights, not finishes.

Grain, freckles, and the tools I did not lead with

The blotchy sky in the failed test is the reason this job sits next to a grain note. Local contrast amplifies whatever small differences exist, including JPEG crumbs and sensor noise. ISO 200 JPEG crumbs are quiet until you ask for them. The grainy photo pass is what I would use if I had to calm a sky I had already overcooked. Better to mask the sky out and never cook it. I did run a very light denoise on a thin band where the hill mask meets the sky, because the blend showed a little mottling in a 40 pixel feather. The band is calm now. The hill texture remains.

Sharpening was the other temptation. The path looked "almost" crisp. A radius near 1 pixel and a modest amount made the path crunchy and picked up single-pixel noise beside it, which reads as scrub. Scrub that was not there is the same class of invention as the fictional tree line. I rejected it. The essay on how to sharpen an image is for edges that exist and have been left soft or have been haloed. This ridge does not have a hard edge. Sharpening a haze line is how you get a drawing of a hill on top of a photograph of a hill.

Enhance photo quality is the tool I use when a file is generally dull, small, or soft and I want one careful pass. I did not start there for the full 5472 frame. A general pass would have spent its effort on faces that were already fine and might have rewritten the freckles. I used the idea of it only as a restrained local-contrast and dehaze on the hill, which I can watch. If the only file had been a phone snapshot of this print, soft and small, the general pass would be the first conversation. We have the camera JPEG. The defect is regional.

Faces got a separate, tiny tone lift so they did not look dim next to a hill that had gained density. It was a curve, masked to skin and shirts, and it did not add clarity. Shirts stayed the colors I remember: a red one, a blue one, a white one that was slightly blue in the shade. I sampled the white shirt in the shade at about 210, 214, 224 and left it. Neutralizing it to a perfect gray would have erased the open shade. Haze work makes people over-neutralize. The shirt is a witness. I kept its cool shade.

A table of the passes, including the ones I threw out

The numbers are samples, not a religion. A different hill will not be 12 levels apart. The columns are what I need in order to stop.

Pass	Where	What changed	Kept
None, the JPEG	Whole frame	Ridge and sky about 12 levels apart	Source
Deblur crop	Hill only	False contour and hint of trees	No
Clarity, no mask	Whole frame	Hair rim, harsh freckles, blotchy sky	No
Clarity, hill mask	Hill	Gap moved to about 22, path about 2 pixels	Yes
Dehaze 15	Distance only	Density in far trees, no cyan	Yes
Dehaze 40	Test on distance	Cyan edges, angry grass	No
Sharpen on the path	Hill	Crunchy 2 pixel line, scrub noise	No
Skin curve	People	A small lift, no added clarity	Yes

The kept rows are three, and they are gentle. The rejected rows are where the picture starts to look "fixed" in a thumbnail and wrong at 100 percent. I judge this file at 100 percent on the path and at a full-frame view on the faces. Both have to survive.

Before I add local contrast to a hazy frame

The order is dull and it keeps the nephew's freckles in the realm of freckles.

• Read the shutter, the aperture, and the ISO, then confirm on a near edge that should be sharp.
• Sample the hazy boundary and a near texture. Write down both spreads.
• If the near edge is sharp, do not deblur the distance.
• Mask people and mask a smooth sky before any clarity or dehaze.
• Raise local contrast only until the structure you remember appears, then sample again.
• Keep dehaze low enough that known colors, such as trees and shirts, do not change family.
• Do not sharpen a line that is only a couple of pixels of recovered contrast.
• If the sky mottles, back off, rather than denoising a wound you did not need to open.

The remembered structure has to be one you can name. "Make it clearer" is not a name. "The switchback path" is a name. "The rock that looked like a chair" would have been a name, and it did not appear, so I did not go hunting for a chair. Named targets prevent a slider from becoming a mood.

A speckled drawing, because clarity eats flat color first

Flat color is where local contrast shows its teeth. I keep a generated drawing that has had coarse speckle laid over it: a round smiling head, a teal body, a teal panel with a yellow circle and three tan bars, a small green shape on a brown floor. The underlying shapes are simple. The speckle is the only fine structure.

If I raise local contrast on that drawing, the speckle becomes the subject and the yellow circle stops looking flat. That is the failure I saw in the picnic sky. The sky was nearly flat, and the crumbs became the subject. The hill had real folds, and local contrast made the folds readable. The drawing lets me show the difference without zooming guests into a child's freckles. It is not a photograph of the picnic and it is not a person. The head is a circle with two dots and a curve.

I also keep a clean version of the same drawing, without speckle, so I can remember that clarity does almost nothing to a truly flat shape except find noise. If a slider "improves" the clean drawing a lot, it is finding something I do not want found. On this job the clean comparison stayed a mental check. The speckled file is the one that matches a hazy sky after too much clarity.

Print size and the weather I refused to replace

She prints 12 by 8 inches for an album page. The file is 5472 by 3648, so a 12 inch long side is about 456 pixels per inch before any crop. We are not short of pixels. Upscaling would not cut haze. Haze is tone, not a shortage of samples. I make the print from the masked edit and I do not add output sharpening in the printer driver. Driver sharpening put a rim on the fence wire in a test strip, a light line on the dark side of the wire, and the wire was already 1 pixel honest.

At 12 inches the path is still subtle. I want it subtle. In the room, in 2016, the path was subtle. A print that shouts the switchback is a map, and she asked for the afternoon. The sky stays pale near the ridge and deeper at the top corner, which is how the frame was exposed, sun off to the side. I did not graduate the sky into a blue postcard. The dehaze at 15 did not do that, and a second pass someone suggested, "just on the sky," would have. The sky is masked out for a reason.

Viewing distance for an album page is the lap, about 14 inches. At 14 inches the blotchy-sky failure is obvious, which is another reason it did not ship. The kept file, at 14 inches, looks like a picnic with a far hill. If you know about the path you can find it. If you do not, you still see a hill and not a smear. That is the correct amount of clarity. More would be a different memory.

Color I left in the day

The light is warm on the table and cooler in the distance, which is normal when haze takes the warmth out of far things. I did not force the hill to the table's warmth. A global white balance toward the table made the hill yellow and the haze looked like smoke from the grill. There was a grill. The smoke is a small gray column at the right, and I left it. Dehaze wanted to turn the smoke into a hard streak. The mask includes the smoke with the sky, so it stayed soft. Smoke and haze are cousins. I am allowed to leave both.

The red shirt was the saturation check. In the original it is a known red, a shirt we still have in a drawer. I held it up. The print matches well enough that I will not nudge reds as a class. Global vibrance had made the shirt neon in a test and had made the dry grass glow. Neon is not clarity. It is a different edit, and it was easy to refuse with the shirt in my hand. Physical references beat memory. The shirt is the reference. The hill has no reference swatch, which is why I stopped at a path I can see and a 22 level gap I can sample, rather than at a feeling of "enough blue."

Files and the sentence on the album page

picnic-2016-original.jpg, picnic-2016-clarity-nomask-rejected.jpg, picnic-2016-hill-only.tif, picnic-2016-12x8.jpg. The text note lists f/8, one four-hundredth, ISO 200, the 12 level sample, the 22 level sample, dehaze 15, and the path width. It also says the nephew's freckles were not retouched. He asked. Teenagers ask. I told him the dirt-speck version existed and that I would not print it. He was satisfied, which is not always the case, and the rejected file remains so I cannot accidentally pretty it later.

The album caption she wrote is about who brought the pie. It does not mention clarity, haze, or a path. Good. The edit is done when the caption can be about the pie. If the caption has to apologize for the sky, I overshot.

How to increase clarity of a photo is a crowded phrase. People use it for focus, for noise, for dull JPEG contrast, and for haze. This file answers one of those. The near wire is sharp. The far hill is veiled. Local contrast and a small dehaze, on the veil only, with faces and sky protected, moved a 12 level gap to about 22 and found a 2 pixel path. That is the job. The pie caption can take it from here.

The neighboring notes, so the next file gets the right one

How to make photos more clear is the wider net, when I am not yet sure the defect is haze. Grainy image is where I send a frame whose "clarity" problem is really noise, or whose clarity pass has already woken the noise up. How to sharpen an image is the last, small edge pass, and it was the wrong first instinct on this ridge. I link them because the search box treats them as one errand. They are not one errand. The fence wire is how I tell them apart, in about a minute, before the slider moves.